INTEGRATING MOLECULAR AND TRADITIONAL SYSTEMATIC TECHNIQUES TO REDEFINE RED ALGAL (RHODOPHYTE) DIVERSITY IN THE BERMUDA ISLANDS

Molecular-assisted alpha taxonomy (MAAT) is a groundbreaking methodology that combines molecular tools with traditional morphological investigations. From studies using these methods, researchers can determine whether specimens with different morphologies are actually one entity exhibiting high phenotypic plasticity or are multiple genetic species with convergent morphologies, an important breakthrough for phycologists since algae are notoriously difficult to identify on morphology alone. Molecular-assisted techniques have also significantly increased the rate of novel species discovery among the algae, especially rhodophytes. From our own biodiversity assessments, we have learned that numerous members of Bermuda’s macroalgal flora have been misnamed, overlooked, or have not been identified as accepted species. Seaweed diversity in the islands overall, as well as the percentage of endemic species, is presumably underestimated. To explore this hypothesis, MAAT methods have been applied to extensive collections of Bermuda seaweeds accumulated since 2010 along with robust phylogenetic analyses incorporating comparative sequence data from around the world. This dissertation examines several results of these efforts. Four genera have been added to the Bermuda flora — Hommersandiophycus, Trichogloeopsis, Yamadaella and Laurenciella, and a number of species uncovered that are new reports for the islands – Centroceras gasparrinii, C. hyalacanthum, C. microacanthum, Liagora mannarensis, Trichogloeopsis pedicellata, Laurencia dendroidea, L.catarinensis and Palisada flagellifera. Eight species new to science have also been described — Helminthocladia kempii, Liagora nesophila, Yamadaella grassyi, Chondrophycus planiparvus, Laurenciella namii, Crassitegula laciniata, Centroceras arcii and C. illaqueans. Over the course of this study, we have accumulated 1875 DNA vouchered specimens collected from 157 sites around the Bermuda platform, as well as 317 specimens from the Florida Keys and 236 from St. Croix in the Caribbean Antilles, all paramount for present and future work. What we have learned already from this small archipelago suggests a overwhelming underrepresentation of diversity in historical records of the islands macroalgal flora, and highlights the importance of generating an accurate baseline dataset for future

Bermuda is an isolated archipelago at the interface of tropical and warm temperate biogeographic zones, making it an ideal location for biodiversity assessments. Situated just over 1000 km east of North Carolina in the Sargasso Sea, Bermuda's climate, water temperature and marine life are heavily influenced by the Gulf Stream (Locke et al. 2013). Because of the seasonal water temperature oscillations (ranging from 18° C in winter to 28°C in summer) the macroalgal assemblage of Bermuda is made up of warm-water tolerant species from the western mid-Atlantic that have persisted during the last ice age, cool-water tolerant Caribbean species carried northward by the Gulf Stream that recolonized the islands since the Pleistocene, and endemic species that have evolved there (Schneider and Searles 1998a). Despite its distant location from North America and tropical summer temperatures, Bermuda's small size presently supports only ca. 450 species of red, brown and green seaweeds, and endemism among these groups is reportedly less than 2% (Schneider and Searles 1998a).
Reports of benthic macroalgae from Bermuda began to appear in the literature in the last half of the 19 th century based upon collections made on just a few voyages, including the "Challenger Expedition" to the islands (Kemp 1857;Dickie 1874;Murray , 1889. In 1917, F.S. Collins and A.B. Hervey produced the first comprehensive marine flora of the islands and distributed the bulk of their specimens in five volumes entitled 'The Algae of Bermuda' as part of Phycotheca Boreali-Americana (P. B.-A., Collins et al. 1912. In 1949, W.R.
Taylor of the University of Michigan collected in Bermuda with his student, A.J.
Bernatowicz (Taylor 1952). Taylor included their data along with that of previous collectors in his comprehensive Marine Algae of the Eastern Tropical and Subtropical Coasts of the Americas (1960). Later, Taylor and Bernatowicz (1969) produced an annotated list of the most common shallow water macroscopic seaweeds of the Bermudas.
From biodiversity assessments like those referenced above, we have learned that a significant number of Bermuda's macroalgal species have either been misnamed, overlooked, or represent dark taxa-specimens that presently aren't identified as a known species. Seaweed diversity in the islands, and conceivably the percentage of endemic species, is likely underestimated. To explore this hypothesis, we have applied molecular-assisted alpha taxonomy (hereafter referred to as MAAT) to extensive collections of Bermuda seaweeds accumulated in August of 2010 and from January to December of 2012. Collections were made from shore access points around Bermuda, as well as by boat, and are representative of a variety of habitats ranging from the intertidal zone to as deep as 36 meters (Fig. 1) Many of the 157 collecting sites indicated in Fig. 1 were visited repeatedly throughout 2012, to capture seasonal variation in macroalgal assemblages with changing sea surface temperatures across the reef platform.
For taxonomists, MAAT is a groundbreaking methodology that combines molecular tools with traditional morphological investigations (see Cianciola et al.

[Appendix A] for a complete review of MAAT and its application to
biodiversity studies). For most groups of eukaryotes, the 5' end of the mitochondrion-encoded cytochrome oxidase I gene (COI-5P) has been established as an ideal, nearly universal, marker that is sensitive enough to differentiate between even closely related species, while remaining conservative enough for comparisons at the taxonomic family-level (Saunders 2005). COI-5P barcode data analysis is an efficient screening tool for identifying genetic species groups in the red algae, and is useful for determining divergences within and among species (based on nucleotide differences in aligned sequences).
In the small representative sampling from Bermuda collections made during the course of this study, the number of genetic species estimated by the COI-5P barcode data is nearly twofold that of currently reported numbers (Wynne 2011) for these taxa in the islands, with 45 presently recognized species in this subsample forming ~82 unique genetic entities in preliminary sequence analyses (Table 1). For example, in the taxon rich red algal order Ceramiales, the data has revealed unreported diversity in the genus Centroceras and among genera in the Laurencia complex. These are discussed in detail in Appendix C (Schneider et al. 2015) and in Chapter 4, respectively. Wrangelia, Spyridia and Dasya, three common western Atlantic genera also in the Ceramiales, are examples of future candidates for robust molecular and morphological analyses. In the genus Wrangelia alone, a single morphological species (Wrangelia cf. penicillata) is shown to be four genetic entities (Fig. 2). This is a classic illustration of diversity that would presumably go undetected without the relatively fast, powerful assessment methods employed in MAAT. The data produced thus far during this study also suggest that the red algal genera Dichotomaria, Chondria, Ceramium, Chrysymenia, Gracilaria and Polysiphonia/Neosiphonia in Bermuda are in need of further molecular and morphological examination.
For phylogenetic assignment of novel species or of taxa that appear to be misplaced based on the results of DNA barcoding, additional molecular markers and alpha taxonomy are employed to confirm the data, ensuring a robust classification.
Standard molecular markers (e.g., the large or small subunit [LSU, SSU] of the ribosomal cistron; the plastid-encoded RuBisCO [rbcL] operon) are often used, but the precise genes used to further investigate targeted taxa are generally chosen based on the taxonomic level the researcher is aiming to resolve, as well as the comparative data available in GenBank (http://www.ncbi.nlm.nih.gov). From these data, phylogenetic trees are constructed with the broadest relevant sampling possible, allowing us to differentiate between undescribed cryptic species and unrecognized additions to the Bermuda flora described from other places in the world. In this way, we can also place Bermuda taxa into a larger biogeographical context.
Cryptic species, in the truest sense, cannot be distinguished using morphological characters, which may overlap partially or entirely across genetic species groups. In other words, regardless of the genetic data that indicate they are distinct entities, morphological examination in the field or lab could lead the researcher to either of two (or more) names that are tied to identical descriptions.
For example, in a new study of the tribe Laurenceiae, or the 'Laurencia complex' (Ceramiales, Rhodomelaceae), the novel genus Laurenciella was ascertained through molecular sequence data as sister to the Laurencia sensu stricto clade.
Though an rbcL sequence divergence of nearly 10% clearly indicated that the two clades differed enough to be separated into distinct genera, no morphological characters could be found to formally support the erection of a new genus (Cassano et al. 2012). Thus, Laurenciella gen. nov. was published with the support of molecular data alone, the first macroalgal genus where this was done.
Pseudo-cryptic species, however, appear superficially identical to congeners, but once molecular distinctness is established, morphological characters can be carefully observed and measured to establish unique identifiers (Maggs et al. 2007).
The 'Centroceras clavulatum complex' (Ceramiales, Ceramiaceae) for example, was considered to be a cosmopolitan complex of cool temperate to tropical red algae, until molecular sequencing of a number of isolates worldwide greatly restricted its biogeography to the Pacific Ocean ). By combining morphological nov. (formerly Nemastoma gelatinosum M. Howe) a taxon that appears remarkably different in gross morphology at different times of the year (Schneider et al. 2011b). This is merely one example that demonstrates the importance of molecular tools to the proper identification of morphologically plastic species.
The scenarios discussed above emphasize the utility of molecular tools for determining whether specimens with different morphologies are actually one entity exhibiting high phenotypic plasticity or are multiple genetic species with convergent morphologies. Molecular techniques are also significantly increasing the rate at which phycologists can identify novel specimens, and as such, an overwhelming number of studies indicate that biodiversity among the algae (rhodophytes in particular) is vastly underestimated (Maggs et al. 2007). For instance, recent analysis of the Sebdeniaceae in the Pacific presented seven undescribed or unknown specimens (Kraft and Saunders 2011). Initially thought to be a species-poor group, the family was originally monotypic with Sebdenia.  added a second genus to the family after molecular and morphological data recovered the novel taxon Crassitegula walsinghamii C.W. Schneid., C.E. Lane & G.W. Saunders from limestone sinkhole pools in Bermuda. After its publication, the biogeographical range of the monotypic genus Crassitegula was extended to the Pacific and a third genus was added to the family, Lesleigha (Kraft and Saunders 2011). Furthermore, our work has uncovered a second novel species of Crassitegula from Bermuda, C. laciniata C.W. Schneider, Popolizio & C.E. Lane (Schneider et al. 2014a) discussed in detail in Appendix B. Molecular sequencing has been paramount to identifying several other taxa new to science in Bermuda, e.g., the novel genus and species Archestenogramma profundum C.W. Schneid., Chengsupanimit & G.W. Saunders (Schneider et al. 2011a)
Molecular tools can also lead us to the discovery of taxa in the study region that have been overlooked in previous assessments. The Indo-Pacific species Liagora mannarensis V. Krishnamurthy  Bodard from the Gulf of Mexico (Schneider et al. 2014a [Appendix B]) and the genus Predaea, not previously reported for the islands, tied to specimens from Mexico identified as P. goffiana D.L. Ballant., H. Ruiz & Aponte (Schneider et al. 2014a [ Appendix B]). To determine whether our specimens truly represent P. goffiana or are novel, comparative material is needed from the type locality of Puerto Rico. Our data has also identified the green alga Codium carolinianum Searles in Bermuda after sequencing material of it from the type locality of North Carolina (Schneider et al. 2014a [Appendix B]).
When a researcher discovers a species that is new to science, she chooses from the collection a specimen to represent the holotype (or type specimen). All future reports of this taxon are compared to the type prior to applying its name. Linking a specimen to the type is the only unmistakable way to validate whether a specimen one has collected is, or is not, the same entity (and in cases of genuine crypsis, comparing molecular sequence data from the specimens in question with the type). Type specimens are essential to phycological investigations because they anchor the classification system by representing the "true" or "standard" example of a described taxon (i.e. species, genus, etc.). The type locality is the location from which the holotype specimen was collected, and serves a similar purpose, especially in molecular phylogenetic studies when it is not possible or practical to gather molecular data from an archival type specimen. Logically, a morphologically comparable specimen identified from the type locality is more likely to be classified correctly than a species with the same name from a disjunct geographical location.
Many of the early phycologists who identified thousands of western Atlantic algal specimens were European, and their bias is evident in the names they applied to specimens collected outside of the eastern North Atlantic. Many species in the Caribbean Sea, for example, are designated by European binomials (see Popolizio et al. 2013 [Chapt. 2]). To determine whether these names were accurately applied in the past for Bermuda and the Caribbean, a comparison must be made between representatives from the collection site and from or near the type locality of the collected and field-identified specimen. For example, historically known specimens of Laurencia obtusa from Bermuda were compared with those from the British Isles type locality. As a result of such comparisons, several species with a type locality in Bermuda presently have molecular sequence data associated with them, and serve as Lane and the green alga Cladophora longicellulata C. Hoek.
It is important to note that together, the cases considered here and in the following chapters represent only a small fraction of the work that is yet to be done to establish a complete flora for the islands using molecular tools. Of the 1875 DNA vouchered red algal specimens collected in Bermuda since 2010, ~675 have been barcoded, and far less than those subjected to additional molecular analysis and alphataxonomic morphological examination. The prospective extent of macroalgal diversity that is yet to be discovered in this small archipelago is staggering. including Antithamnionella bermudica sp. nov. Phycologia 36:12-23. Schneider, C.W. & Searles, R.B. 1997b. Notes on the marine algae of the Bermudas.

INTRODUCTION
Despite the heavy influence of the Caribbean on the marine flora of Bermuda , the taxonomic preconceptions of early workers had a major influence on seaweed floristics for the archipelago. These 19th century botanists often used familiar European binomials when cataloging or reporting on the Bermudian flora (Kemp, 1857;Dickie, 1874;Murray, , 1889   and the western Atlantic (Wynne, 2011).
We have noticed that Helminthocladia calvadosii in Bermuda has obvious habit differences with specimens from the eastern Atlantic, and as such, we compared anatomical characteristics, as well as COI-5P gene sequences of Bermuda specimens, to measurements and sequences in GenBank from specimens collected in Europe. We have also extended our investigations to congeners with overlapping characteristics to determine whether our specimens of H. calvadosii sensu auct. from Bermuda require taxonomic action.

Standard Methods
Collections were made in shallow water (0-2 m) and individuals were pressed fresh onto herbarium paper as permanent vouchers. Small fragments were excised prior to pressing, part desiccated in silica gel and the remaining preserved in 4-5%  . The P. B.-A. specimen used in this study is from the exsiccata in CWS' personal herbarium.
The extra-Bermuda collections were processed at the University of New Brunswick following established field and vouchering protocols .
Vouchers are deposited in UNB and all of the pertinent metadata are publicly available in the dataset HELMN01 on the BOLDSYSTEMS web site (www.barcodinglife.org) and summarized here (Table 1).

Molecular Methods
Silica dried samples for DNA analysis were ground in liquid nitrogen and stored at -20°C (Table 1). For Bermudian collections, DNA was extracted from 100 µl ground material using the DNA extraction buffer from Saunders (1993) followed by incubation at 23˚C for 1 hr and then incubation on ice chips for 20 min. Samples were spun at 10,000xg for 10 minutes and the elute was transferred into the extraction column of the Sigma-Aldrich (St. Louis, MO) GenElute Plant Genomic Miniprep Kit.
The remaining protocol was followed according to manufacturer's protocol. DNA extraction of non-Bermuda collections followed published protocols .
For Bermudian collections the COI-5P region was PCR amplified with the Takara Ex-Taq DNA polymerase kit (PanVera, Madison, WI, USA) in an Eppendorf AG Mastercycler epGradient thermal cycler (Hamburg, Germany). Amplified DNA was treated with Sigma-Aldrich GenElute PCR Clean-Up Kit following the manufacturer's protocol and the purified PCR product was sequenced at the Rhode Island Genomics and Sequencing Center using the ABI 3130xl genetic analyzer. For the non-Bermuda collections the COI-5P was amplified following published protocols that do not require subsequent cleaning of the product  with the sequencing outsourced to Genome Québec (www.genomequebec.com). The actual primer pair used with each specimen is available in the dataset HELMN01 on the BOLDSYSTEMS web site and with each entry at GenBank (Table 1).
Fourteen COI-5P sequences from representative Helminthocladia spp., including those available through GenBank and newly determined here, were included in an alignment with 14 sequences (outgroup) from the closely allied species Cumagloia andersonii . Sequences were first aligned in Geneious Pro on a MacPro (OS X version 10.6.8)    with BIONJ used to designate the starting tree, best of nearest-neighbor interchange (NNI) or subtree pruning and regrafting (SPR) branch-swapping options, and with the tree topology, branch lengths and substitution rates optimized. Data partitioning was not implemented. Branch support was estimated using 500 bootstrap replicates.

RESULTS AND DISCUSSION
Helminthocladia 'calvadosii' has been found from January to April in intertidal and shallow subtidal habitats on both the southern and northern shores of Bermuda Is., as well as the northwestern coast of St. George's Is., at rocky sites experiencing moderate to heavy wave action. Although previously recorded as abundant from a number of Bermuda's bays (Taylor & Bernatowicz, 1969), at present we have found this species to be infrequent to rare except at the Spanish Point site where it remains abundant. The winter-spring collections were found when seawater temperatures in Bermuda average from 18-20°C, the cooler end of the warm temperate biogeographic range. Of the sixteen currently accepted species of Helminthocladia (Guiry & Guiry, 2012), most are found in warm temperate seas in the spring-summer months (O'Dwyer & Alfonso-Carrillo, 2001), and a few are found in tropical seas (Guiry & Guiry, 2012 (Searles & Lewis, 1983). It differs from Bermudian H. 'calvadosii' by its small size (to 5 cm tall), overall habit and lax medulla, production of descendant rhizoids from sterile cells, and a strongly developed involucre around the gonimoblasts (Searles & Lewis, 1983). The two Australian species, H. densa and H. dotyi, are repeatedly subdichotomous and thus more densely branched than the Bermuda specimens, the latter being a small species, 2-7 cm tall, with a massive base (Womersley, 1965;. Helminthocladia densa has barely enlarged terminal cells on cortical fascicles (Womersley, 1965), unlike the specimens collected in Bermuda. Both of these species lack the heavy investment of adventitious branches around their axes, and neither would be confused with the Bermuda collections.
Evaluating the COI-5P barcode sequences, Bermudian specimens form a distinct genetic species group that resolves deeply among the other included species of Helminthocladia being most closely allied ( Fig. 1 (Kemp, 1857;Murray, 1889). Setchell (in Collins et al., 1915) alerted    , H. calvadosii has been reported for the western Atlantic from Dominica (Taylor, 1969), Venezuela (Díaz-Piferrer, 1970) and Brazil (Guimarães et al., 1990). The only of these reports with morphological and anatomical data is that from Brazil (Guimarães et al., 1990), and it shows that these South  (Grant Nos. 2002-34438-12688, 2003-34438-13111 and 2008, and the     . Its members are widely distributed in tropical and temperate seas, but exhibit the greatest diversity in warm waters (Huisman 2002. Lin et al. (2015) have recently recognized six families in the order, separating out three families from the Liagoraceae by resurrecting the Nemaliaceae and creating two monogeneric new families, the Liagoropsidaceae and the Yamadaellaceae.
The distinctive anatomical and reproductive characteristics of the Liagoraceae sensu lato include the presence of multiaxial thalli with a subdichotomously or trichotomously branched filamentous cortex, and post-fertilization development where the gonimoblast initiates directly from one or both cells of the divided carpogonium (Huisman 2002. Traditionally, genera and species have predominantly been recognized based on distinguishing reproductive characteristics such as the size, shape and origin of the carpogonial branch, the nature of the gonimoblast (compact vs. diffuse), the presence (or lack thereof) and arrangement of the sterile filaments associated with the carposporophyte, as well as spermatangial development (Kraft 1989, Huisman 2002, Lin et al. 2015. In addition, vegetative characters such as branching patterns, degree of calcification, and cortical cell shape and size are common identifiers, albeit generally less salient. Given that a great number of species in the Liagoraceae have remarkably similar morphologies, the group has suffered a long history of taxonomic ambiguity (Lin et al. 2013). To remedy this, modern diagnostics of generic (and specific) placement integrate DNA sequence analysis with the examination of carposporophyte development and other characters, resulting in a stronger taxonomic organization and a more accurate representation of biodiversity.
Though members of the Liagoraceae sensu lato were significantly consolidated following several morphological studies in the 20 th century (e.g., Abbott 1990aAbbott , 1990b, the increasing use of molecular sequence data in systematic research over the past decade has resulted in the descriptions of several new genera and species

Standard methods
Collections were made in shallow water (0-3 m) or via scuba (0-23 m) and site locations were taken using a Garmin™ eTrex H (Olathe, Kansas, USA). A portion of each specimen used for DNA analysis was then dried on silica gel and the remainder of the thallus was pressed onto herbarium paper as a permanent voucher. Selected fragments were preserved in 4-5% Formalin in seawater for anatomical study. Herbarium abbreviations follow the online Index Herbariorum <http://sweetgum.nybg.org/ih/> and standard author initials are from .

Molecular methods
Specimens used in molecular analyses are recorded in Table 1 Freshwater and Rueness (1994). In some instances, the 3' portion of the rbcL was alternatively amplified and sequenced using the primers (TLF4 with rbcLrevNEW) and amplification profile presented by . The LSU marker was amplified and prepared for sequencing using the primers and protocols outlined by Harper and Saunders (2001) and  Based on these groups, and with comparative data available from GenBank, one specimen from each species and/or geographic location was selected for phylogenetic analysis using rbcL and LSU sequences. The most variable portion of the 3' "Z" fragment (Saunders and Moore 2013) of the LSU marker was missing from several GenBank specimens (~215 bp). The remaining sequence data in this fragment were highly conserved, between 97.1% and 99.8% identical across all genera in the family.
Thus, the alignment was trimmed of the "Z" fragment and the more variable region representing the "X" and "Y" fragments (Saunders and    and run with four parallel chains (three heated + one cold) with branch lengths optimized during the run for one million generations. The initial 2500 trees were discarded as the burn-in, and posterior probabilities were estimated based on the remaining trees. The LSU analysis was conducted with the same parameters as rbcL, but with two million generations. Stationarity was attained after the first 500,000 generations (burnin = 5000 trees). Both the rbcL and LSU gene trees include members of the closely related Scinaiceae and Galaxauraceae families as outgroups. All trees (Figs 1, 2 and S1) were manipulated for presentation using FigTree software (http://tree.bio.ed.ac.uk/software/figtree/).

Molecular observations
Our DNA barcode analyses included 105 sequences from members of the Liagoraceae and Yamadaellaceae; 78 of these were from Bermuda collections. The final COI-5P barcode alignment consisted of 541 base pairs, of which 40.5% were informative characters. This tree (Fig. S1) was used to initially align genetic species groupings  For the LSU analysis, the ML result is shown, with bootstraps and Bayesian posterior probabilities respectively at the nodes (Fig. 2). Many of these support values are noticeably problematic, despite the generic placement of specimens across the tree being mainly congruent with that of rbcL. The LSU sequence alignment indicated that intergeneric distances of 0.8-7.5% were remarkably smaller than in COI-5P and rbcL due to the highly conserved nature of this gene, potentially resulting in weak support values for the topology recovered. Some comparative sequence data from GenBank were not as reliable for this gene compared with rbcL, therefore the absence of data needed to resolve the relationships among liagoracean taxa may also contribute to the lack of support, especially toward the backbone of the tree. However, we chose to present the LSU analysis in spite of the difficulty resolving a well-supported tree, because the data confirm several important findings. Most notably, the LSU data were essential for identifying an unknown specimen grouping with the genus Liagora in the previously sequenced genes. The LSU sequence was identical to a GenBank sequence of L. mannarensis from western Australia. Our COI-5P data showed that we collected a number of specimens most closely related to Yamadaella caenomyce (9.5% distance) from Hawaii. and LSU sequence data confirmed with full support that Bermuda specimens belonged in the same genus, but were distinct from the Indo-Pacific generitype. Also, although its position within the Liagora clade is uncertain in the LSU analysis, the data provide further evidence that Liagora ceranoides, the species known in Bermuda prior to this publication, does not appear to be present in Bermuda waters. Instead, the collections represent a novel taxon that is closely related to L.  Fig. 3; data in Table 1).

Remarks:
Only two known representatives of this rare species have been collected, each at the same location in 2009 and 2012, several miles offshore at a depth of 15-16 m, attached to coral rubble. The collections were initially fieldidentified as Liagora ceranoides, but our COI-5P and rbcL sequence data later showed that the 2012 specimen was genetically distinct from all other Liagora specimens collected and sequenced from the islands. Moreover, this specimen did not group with any other liagoracean species from ours or the comparative data, instead resolving on its own branch, basal to the Liagora clade in both analyses. A publicly available LSU sequence of L. mannarensis from Western Australia was identical to LSU data from our Bermuda specimen (Fig. 2). An examination of the morphology and reproductive characters of our plant agreed with the description of collections from Western Australia (Huisman 2002), as well as the type from India (Krishnamurthy and Sundararajan 1985). Clearly, L. mannarensis is cryptic with was has been called L.
ceranoides in Bermuda and the only way one can reliably distinguish them (other than using molecular markers) is to observe the production of gonimoblast initials from both the proximal and distal cells of the transversely divided carpogonium. According to Huisman (2002), all other known species of Liagora generate gonimoblast filaments exclusively from the distal cell. Carpogonial branches in our Bermuda specimens are remarkably small (Fig. 4), making observation of the reproductive characters important for characterizing the species difficult, however we were able to locate a fertilized carpogonium where longitudinal division of both initial daughter cells was evident (Fig.5). The Bermuda specimens appeared to have more weakly developed involucral filaments ( Fig. 7) compared with Western Australian specimens (Huisman 2002). Interestingly, Krishnamurthy and Sundararajan (1985) note their Indian specimens are present only in fall and winter, disappearing abruptly from the flora in spring; we collected both of our well-developed Bermuda plants in March, suggesting the species may have a similar growing season in the western Atlantic.
This study represents the first report of L. mannarensis in the Atlantic Ocean, and is one example of several Indo-Pacific genera or species in the Liagoraceae and Yamadellaceae discovered in this region.

Liagora nesophila Popolizio, C.W. Schneider et C.E. Lane sp. nov. (Figs 8-14)
Description: Thalli to 14 cm, whitish to brownish rose, usually with dark red tips, lightly to moderately calcified. Branching dichotomous to many orders, with obtusely angled dichotomies at the branch apices, and some with adventitious branches present on main axes that at times appear pinnate. Axes 0.  Table 1 as CWS/CEL/TRP 12-10-10, TRP 12-28-3, TRP 12-32-9, TRP 12-37-7, TRP 12-51-3 and is reasonable to assume that non-sequenced specimens from the islands identified as such by earlier workers (see ) are all L. nesophila, effectively removing L. ceranoides from the flora, and providing a geographic distinction for these two species. Collections: Specimens listed in Table 1.

Remarks: This represents the first report of Trichogloeopsis pedicellata from
Bermuda. Based upon the carpogonial branch shape and the unique sterile filament development associated with the carposporophyte, Abbott and Doty (1960) transferred Liagora pedicellata to their new genus Trichogloeopsis. In habit, Bermuda collections are similar to Gloiocallis dendroidea from Bermuda and other western Atlantic sites.
These two species are situated in the same clade in our phylogenetic analyses ( Fig. 1) with a genetic distance of 10.8% for rbcL. Molecular sequence data from our specimens plainly illustrate that T. pedicellata from Bermuda corresponds to sequences from the Bahamas Island chain, the type locality of this species (Fig. 1).
Trichogloeopsis is characterized by the presence of short sterile filaments that issue from below the gonimoblast during carposporophyte development, these referred to as descending 'gonimorhizoids' (Abbott and Doty 1960) or 'sterile rhizoids' (Vélez-Villarmil et al. 2000). All species of Trichogloeopsis display this trait, but only T. pedicellata possesses pedicels, the particularly long bearing or stalk cells at the base of the carpogonial branches (the namesake of the western Atlantic species). Our collections from Bermuda demonstrate both of these distinctive characters (Figs 16 and 17). Surprisingly, Bermuda specimens commonly exhibit multiple carpogonial branches originating from a single basal supporting cell (Fig. 19 Etymology: Named for Roger "Grassy" Simmons whose local expertise, support and friendship were invaluable during the first author's field collection year in Bermuda, and who introduced her to the site where this species was first discovered.
Distribution: Thus far, only known from the type locality and nearby intertidal sites on the southeastern shore of Bermuda Is., from Grape Bay to Devonshire Bay.

Remarks:
The genus Yamadaella was described by Abbott (1970) (Abbott 1970, figs 3-5). Finally, we have observed only dioecious gametophytes in the many Bermuda specimens rather than strictly monoecious gametophytes for Indo-Pacific populations (Abbott 1970, Lin et al. 2015. Yamadaella grassyi exhibits inflated, wedge-shaped terminal cortical cells that are typical for the genus (Fig. 21) and allow it to be easily distinguished from the other genera reported here.
Originally thought to be restricted to the Indo-Pacific region, Yamadaella caenomyce was later reported by Wynne and Huisman (1998)  Collections: Specimens listed in Table 1.
Distribution: Reported as pantropical, but not corroborated by molecular data.

Remarks:
The first report of Liagora valida from Bermuda was made by Kemp (1857) and this was followed by numerous reports, as it is common during summer months. In the last decade, this species was moved to a new genus, Titanophycus validus has also been reported from Australia, Taiwan, Japan, and Hawaii, indicating that as presently circumscribed, it is a pan-tropical species. However, additional molecular data is needed for specimens collected in the Indo-Pacific region to validate this. Lin et al. (2011) remark that the specimens they collected in Taiwan do not correspond molecularly, despite reports of T. validus from the region. They also observed that sequences from Hawaii grouped with their T. setchellii, suggesting possible crypsis. Our phylogenetic analyses agree with these findings, illustrating that "T. validus" from Hawaii is distinct from Bermuda material and collections from near the type locality, with 6.9-7.1% genetic distance for COI-5P (Fig. S1). Moreover, T.
validus from Western Australia exhibits 0.5% distance from Bermuda specimens in the portion of the highly conserved LSU gene sequenced in this study ( Collections: Specimens listed in Table 1.
Distribution: Bermuda, Florida, Bahamas, Caribbean, Brazil, Mariana Is. Remarks: Taylor (1960) was the first to report Liagora mucosa M. Howe from Bermuda, and later, this species was reduced to synonymy of L. dendroidea (P. Crouan et H. Crouan) I.A. Abbott (Abbott 1990b). Subsequently,  transferred L. dendroidea to Ganonema based on reproductive characters associated with the carpogonial branch and spermatangial arrangement.
When a recent phylogenetic analysis rendered Ganonema polyphyletic, a new genus,

Gloiocallis, was established for G. dendroideum (Lin et al. 2014). At present,
Gloiocallis is monotypic and is sister to Trichogloeopsis. It is characterized by lacking the descending gonimorhizoids associated with Trichogloeopsis, and by possessing carpogonial branches that are produced terminally or laterally from the cortical filaments (compared with Ganonema in which they originate in a special accessory branch system). Our Bermuda specimens' morphologies conform to those described for G. dendroidea by Huisman (2002), Ballantine and Abbott (2006) and Lin et al. (2014). Abbott (1990a) remarked on the similarity between G. dendroidea and T. pedicellata, stating they were "equally gelatinous and pinnately-paniculately branched." She also noted that T. pedicellata was more common in the Caribbean waters she had examined them from than the former species was. Conversely, in Bermuda G. dendroidea is far more common than T. pedicellata.

When Ballantine and Aponte (2002) transferred Liagora dendroidea to
Ganonema, they mentioned that polycarpogonial branches were occasionally produced in this species. As evidenced in this study, this is yet another feature shared by Gloiocallis and the closely related genus Trichogloeopsis. At present, the only conspicuous morphological character used to distinguish the two is the lack of descending gonimorhizoids in the latter species. It is worth noting that the intergeneric distance between T. pedicellata and G. dendroidea (10.8% for rbcL) is similar to values found for interspecific relationships among the Liagoraceae. For example, the genetic distance between H. pectinatus and other members of Hommersandiophycus (9.2-10.8%) or between Liagora viscida (Forsskål) C. Agardh (generitype) from Spain and L. ceranoides (10.7-11.0%), L. nesophila (9.4%), L. albicans J.V. Lamouroux Collections: Specimens listed in Table 1.
Description: Thalli to 20 cm high, brownish-rose to dark red, with light to moderate calcification. Branching alternate to irregularly alternate, sometimes opposite above, axes 1.0-3.0 mm diam. Ultimate branchlets often pectinate or secund.  Table 1.  (Fig. 29). In addition to its restricted geography, the pectinately arranged branchlets often found on well-developed specimens of H. pectinatus (Fig. 26) differentiate the western Atlantic species from its Indo-Pacific congeners.

Concluding remarks
The Indo-Pacific region is the most speciose region for the Liagoraceae (Lin et al. 2013), with 25 genera and 152 species presently reported .
With records of 13 genera and only 24 species, the warm temperate to tropical western Atlantic region is comparatively depauperate . However, we have demonstrated that the diversity of this group in the western Atlantic, and Bermuda in particular, has been undervalued. The number of liagoracean genera in Bermuda have nearly doubled from the five reported prior to our investigation ; we have determined that at least nine genera are present. Furthermore, we have extended the range of the Caribbean species Trichogloeopsis pedicellata to its northernmost outpost in Bermuda, and confirmed that specimens previously described as Ganonema dendroidea in Bermuda are equivalent to Gloiocallis dendroidea from Brazil. We transferred Liagora pectinata with a type locality in Bermuda to the recently segregated Hommersandiophycus, thus expanding the range of an Indo-Pacific genus to include the Atlantic Ocean. Liagora mannarensis, previously known only from the Indian Ocean, is now recognized for the first time in the Bermuda flora and the Atlantic Ocean. Moreover, we have described two novel species, Liagora nesophila and Yamadaella grassyi. The latter is the first report of the genus Yamadaella in Bermuda. Convincing molecular evidence from this study suggests Liagora ceranoides is not present in the waters surrounding Bermuda.        showing extensive sterile filaments . Scale bar = 100 µm. 14.

INTRODUCTION
The 'Laurencia complex' is a diverse sub-grouping of red algae in the Rhodomelaceae family (tribe = Laurencieae Schmitz 1889) found in tropical and temperate seas worldwide, and contains 208 currently accepted species (Guiry & Guiry 2015). The group has received substantial attention from phycologists in the past several decades due to its convoluted taxonomy. Within the complex (and in addition to Laurencia sensu stricto) the genus Osmundea has been resurrected (Nam et al. 1994), Chondrophycus has been elevated to generic status (Garbary and Harper 1998), and three new genera have been named-Palisada, Yuzurua, and most recently, Laurenciella (Nam 2007;Martin-Lescanne et al. 2010;Cassano et al. 2009).
Members of the 'Laurencia complex' are known to exhibit considerable morphological plasticity, which can make them challenging to diagnose from a morphological point of view. However, each of the genera has been shown to be wellsupported monophyletic clades in multiple publications, except for the taxon-rich genus Laurencia (145 spp.). This group is in need of extensive molecular and morphological analysis and, likely, reorganization.
As a genus, Laurencia has already been significantly modified by molecularassisted alpha taxonomy (MAAT). Erected by Lamouroux (1813: 131-132)  recently been added to the complex for a molecularly distinct clade that is entirely morphologically cryptic with sister genus Laurencia sensu stricto (Cassano et al. 2012b).
Members of the tribe Laurencieae are distinguished from the closely related genus Chondria by both vegetative and reproductive characters. Chondria displays five pericentral cells per axial cell (mostly obvious only at the apices); genera in the Laurencia complex produce either four or two (Womersley 2003). Spermatangial plates form in Chondria; alternatively, development is from trichoblasts located within apical pits in Laurencieae (Nam 1999).
Several workers in the past two decades have provided a phylogenetic basis for reorganization of the Laurencia complex within a global context , Gil-Rodriguez et al. 2009, Martin-Lescanne et al. 2010. Comprehensive surveys of the group using molecular tools are less common for discrete localities like Bermuda, but the results are particularly interesting when compared to historical accounts of the islands' flora. Reports of Laurencia in Bermuda begin to appear in the literature in the mid 19 th century. The earliest reports of the Reverend Alexander Ferrie Kemp (1857) included L. obtusa (Huds.) J.V. Lamour. and L. papillosa C. Agardh (now Palisada perforata (Bory) K.W. Nam). Laurencia obtusa (type locality = England) was subsequently recorded from the islands by workers over the next century , Dickie 1874[as P.B.-A. 42:2092, , Howe 1918, Tandy 1936, Bernatowicz 1952) followed by additional accounts in the 20 th century. Laurencia obtusa var. crucifera Kütz. was reported by Dickie (1874), and var. gracilis (C. Agardh) Zanardini by .  reported L. paniculata J. Agardh from Bermuda and this species is presently regarded as a junior synonym of L. obtusa. Rein

Standard methods
Collections were made in shallow water (0-3 m) or via scuba (0-23 m), and site locations were taken using a Garmin™ eTrex H (Olathe, Kansas, USA). A portion of each specimen used for DNA analysis was dried on silica gel and the remainder of the thallus was pressed onto herbarium paper as a permanent voucher. Selected fragments were preserved in 4-5% Formalin in seawater for anatomical study. Sections were mounted in 30% corn syrup with acidified 1% aniline blue in a ratio of 20:1 with a few drops of Formalin as a medium preservative. Live specimens chosen for DNA analysis were photographed using a Canon Powershot s90 digital camera (Canon Inc., Tokyo, Japan) and dried herbarium specimens were scanned on an HP 309a Photosmart Premium scanner (Hewlett-Packard Company, Palo Alto, California, USA).

Molecular methods
Specimens used in molecular analyses are recorded in Table 1  COI-5P sequences from representative species within the Laurencia complex, including those available through GenBank and newly determined here, were included in an alignment using the MUSCLE (multiple sequence comparison by log-expectation) alignment program in Geneious (v. 6.1.8 available from http://www.geneious.com). To visually characterize genetic variability among specimens, the UPGMA clustering algorithm was applied to the COI-5P alignment (180 specimens, 458 sites) with Tamura-nei-corrected distances (default setting). The resulting tree was used to demarcate genetic species groups. Based on these groups, and with comparative data available from GenBank, one specimen from each species and/or geographic location was selected for phylogenetic analysis using rbcL sequences. The best models of evolution for the individual gene region rbcL (85 taxa, 1217 sites) were determined in jModelTest 2 (volume 2.1.5; . The selected phylogenetic model (GTR + I + G in both instances) was used to complete both maximum likelihood (ML) and Bayesian analyses for each gene. The rbcL maximum likelihood phylogeny was estimated using the RAxML graphical user interface (Silvestro and Michalak 2012) with branch support calculated using 1000 bootstrap replicates. Bayesian analysis of rbcL was conducted in MrBayes v.3.2.2  and run with four parallel chains (three heated + one cold) with branch lengths optimized during the run for one million generations.
The initial 3650 trees were discarded as the burn-in, and posterior probabilities were estimated based on the remaining trees. The rbcL gene tree includes members of the genera Chondria and Bostrychia as outgroups. All trees (Figs 1, 2) were manipulated for presentation using FigTree software (http://tree.bio.ed.ac.uk/software/figtree/).

RESULTS AND DISCUSSION
Our DNA barcode (COI-5P) analysis included 180 sequences from members of the Laurencia complex; 170 of these were generated for this study from specimens collected in Bermuda, the Florida Keys, and St. Croix, USVI. Ten additional sequences were downloaded from GenBank, and represented several specimens collected in the Pacific (mainly Hawaii) and two species of Osmundea from France.
The final COI-5P alignment consisted of 458 base pairs, of which 35.2% were informative characters. Intergeneric distance values for the five genera represented in our collections (all members of the 'Laurencia complex') were 8.3-12.9%. The barcode analysis was primarily used to organize our large dataset into 13 distinct genetic species groups (Fig. 1). For further generic and specific identification, the rbcL gene was far more informative, especially given the extent of molecular studies using this gene in the past decade.
To place them into a phylogenetic context, a representative of each of the previously determined genetic groups was included in a single gene (rbcL) analysis along with a diverse representation of taxa designated to the complex mined from GenBank (Table 1). Tree topologies from maximum likelihood (ML) and Bayesian analyses of rbcL were consistent; the phylogeny is illustrated by the ML tree, with bootstraps and posterior probabilities indicated at the nodes (Fig. 2). Phylogenetic associations among genera in the tribe conformed largely to those published by Cassano et al. (2012b). Our analysis similarly depicts the sister relationship of Laurencia and Laurenciella, with Yuzurua resolving basally to this subgroup, but as in the aforementioned papers' results, only the Bayesian posterior probabilities show strong support for these two relationships. Whereas the phylogeny in Cassano et al. suggests Palisada is the basal genus in the complex, ours instead indicates a wellsupported basal grouping of the Osmundea-Chondrophycus clade.
Intergeneric rbcL distances in the complex were 8.7-14.5%, similar to the COI-5P distance results. For genera present in our collections, the greatest interspecific variation occurs among Chondrophycus (2.1-8.9%) and Laurencia (2.5-9.0%). That the higher range of these values overlap with intergeneric distances in the complex is indicative of the need for greater taxon sampling (including type specimens or specimens from type localities) and robust molecular analyses in these groups, and it is not unreasonable to expect that future studies will propose additional generic diversification within the Laurencia complex. Notably, specimens we have identified as having "red" habits in our collections group only within the L. dendroidea clade, along with green and purplegreen morphs of this species. All specimens with thalli we have described as 'green with pink tips' in our field notes all fall within the L. catarinensis clade. It is likely that Howe (1918) was examining collections of L. catarinensis when describing 'L.
obtusa' from the islands as "subglobose tufts" that are "often greenish with red tips," features that appear to characterize the habit of this species based on our observations and genetic sequences.
The presence of Laurencia intricata in Bermuda has been verified with molecular data. Sequence data shows that this species, first reported for the islands by Howe (1918), is indeed conspecific with specimens from the Caribbean Antilles, the type locality of this species. We have discovered a specimen of L. intricata among Hervey's archived material collected at Heron Bay, Bermuda (P. B.-A no. 1937, as Laurencia tuberculosa J. Agardh). This exsiccate has specimens of this number representing at least two heterotypic members of the 'Laurencia-complex'. Early workers moved all P. B.-A. no. 1937  reported L. obtusa var. gelatinosa (Desf.) J. Agardh from the region, a name that has since been placed in synonymy with L. microcladia.  describe their collections (reported as L. obtusa var. gelatinosa) as a "low and slender form of exposed rocky shores," archetypal characteristics of this species. Howe (1918) also notes that intertidal collections of L. microcladia from Bermuda are included in "Phyc. Bor.-Am. 1888, as L. obtusa var. gelatinosa." In his account, Børgesen highlights the pyramidal shape of the thallus, nature of verticillate branching with swollen, clavate branchlets, and the tendency of this species to exist in particularly exposed habitats. His illustration of this species, which he reports from the US Virgin Islands, bears a remarkable resemblance to our specimens of L. microcladia from St.
Croix, USVI (Fig. 3).  provides the first comprehensive description of L. microcladia for the western Atlantic region, which fits observations of our specimens from both Bermuda and St. Croix with morphological overlap in all characters for which data have been reported (Table 2) Sentíes et al. (2001) from the Mexican Caribbean, and described to be "entirely in concordance" with Japanese material examined by Saito (1964) and without "significant differences" from Yamada's (1931)  venusta, sparse, verticillate branching is emphasized. Laurencia microcladia is densely branched and possesses verrucose branchlets that are clustered in a manner that creates a whorled appearance (Fig 4), however, branching overall is more irregular than verticillate (Fig 7). Additionally, though the observed diameters for tetrasporangia and cystocarps in the two species overlap, the upper limits of these tend to be larger in L. microcladia than in L. venusta, especially compared with Caribbean specimens of the latter species (Table 2).
Our rbcL sequences from specimens of Laurencia microcladia are almost identical to those from L. venusta collected in Caribbean Mexico, with only 0.2-0.4% divergences between the sequences. Specifically, 2 of 1217 base pair were different between our St. Croix specimen and the GenBank data for L. venusta from Caribbean Mexico; 5 differences were present between both Caribbean specimens and Bermuda specimens. It is quite possible that the entities shown in the previous analyses are genetic variants of L. microcladia, a name that would be precedent over L. venusta by more than half a century. Alternatively, the two may represent cryptic species from different ocean basins, but given the near genetic equivalence of these specimens as clearly demonstrated by our data, we think this is unlikely. Molecular sequence data for specimens of L. venusta from Japan will be required to bring clarity to this puzzling matter.
Our specimens of Laurencia microcladia are also closely related in rbcL The molecular sequence data we have produced verifies that Palisada perforata in Bermuda is conspecific with specimens from the type locality in the Canary Islands, Spain. The species is the most common member of the genus Palisada in our collections, and is predominantly intertidal with a few exceptions from shallow subtidal sites in the Florida Keys and a single subtidal collection from a shallow (3)(4) m) site off Bermuda's north shore. Originally known from the Indian Ocean, Palisada flagellifera has been shown to be pan-tropical with reports from Cuba (Areces et al. 2003), Brazil (Fujii et al. 2006)  corallopsis (Howe 1918) until Littler and Littler argued that it was much smaller in habit and had smaller surface cells than P. corallopsis .
The Littlers were the first in modern times to resume using the name L. cervicornis for Caribbean material .
Specimens field-identified as Palisada corallopsis from Bermuda and the Florida Keys are genetically variable, as shown by both the COI-5P and rbcL analyses (Figs 1, 2). For the latter gene, one genetic entity from Bermuda is closely related (0.6% divergence) and likely conspecific with specimens from Caribbean Mexico also identified as P. corallopsis. The other is sister to specimens we collected in the Florida Keys (2.4% divergence). Further study is needed to determine whether either of our Bermuda variants of P. corallopsis fits the protolog of L. cervicornis from Key West.
Moreover, we must determine whether our collections from the Florida Keys, which also display genetic variation when compared with Caribbean specimens' sequence data, can be morphologically tied to this long-synonymized name. Our COI-5P barcode investigation indicated a group within the Palisada clade that was distinct from reported Bermuda species. The rbcL phylogeny allies this grouping with P. flagellifera from Brazil and the Canary Islands, Spain, but comparative molecular data is not available for specimens from the Indian Ocean type locality of this species. The present report represents the first record of P. flagellifera in the Bermuda Islands. Recently, sequence data has shown that Palisada iridescens is closely aligned to Yuzurua, and a new combination, Y. iridescens is proposed in a forthcoming publication (Sentíes et al. in press). Our data agree with this transfer, as previous records of this species in Bermuda (Schneider and Lane 2007, as C. iridescens) group with the Yuzurua clade in our COI-5P barcode analysis (Fig. 1). Our analysis has also resolved a unique sequence from a Bermuda specimen in this clade sister to Y. poiteaui (4.3% divergence). Thus far, we are only able to confirm a single specimen with sequence data representing this entity, and will require additional collections with matching sequences before we verify that these represent a unique, third species in the genus. For the time being this taxon will be regarded as Yuzurua sp. Hawaiian Islands, group with our Bermuda species in COI-5P distance analysis. The rbcL data shows that the new species is closely related (0.8% divergence) to an undescribed specimen collected from Flower Garden Banks in the Gulf of Mexico (Fujii et al. 2006). These are arranged sister to the rest of the Chondrophycus clade, which consists C. cf. undulatus and several undescribed species from New Caledonia, and C. tronoi from the Philippines. Saito (1967) split the genus Laurencia into two subgenera, Laurencia and Chondrophycus, on the basis of tetrasporangial development relative to the axis (parallel or right-angle) and presence or absence of secondary pit connections between cortical cells. He later determined these features were unreliable since some species possessed features of both subgenera, namely a parallel arrangement of tetrasporangia and no secondary pit-connections (Saito 1982). Species exhibiting this combination of characters would later be attributed to Osmundea (Nam 1994). Ultimately, though, Garbary and Harper (1998) elevated the subgenus Chondrophycus to generic rank based on morphological cladistics, including the absence of secondary pit connections.

Reports of the species currently known as
The genus was shown to be polyphyletic in a molecular analysis by Abe et al. (2006) with disjunct clades sister to either Osmundea or Laurencia. The species in the latter clade were transferred to Palisada (Nam 2007), a genus defined by palisade-shaped epidermal cells, and distinguished from Chondrophycus by differences in tetrasporangial development.

Nam (1999) proposed an infrageneric classification scheme for
Chondrophycus including four subgenera-Chondrophycus, Kangjaewonia, Palisada and Yuzurua. Members of the subgenus Chondrophycus (containing C. cartilagineus (Yamada) Garbary & J.T. Harper as the type, as well as several others) exhibit rightangle tetrasporangia, but lack secondary pit connections. For species displaying both right-angle arrangement of tetrasporangia and secondary pit connections, Nam (1999) circumscribed the subgenus Yuzurua, which has since been elevated to generic status and is irrefutably distinct in phylogenetic analyses from Chondrophycus (Martin-Lescanne 2010). Furthermore, Nam (1999) defined the section Parvipapillatae for members of subgenus Yuzurua demonstrating epidermal cell projections at branchlet apices in transverse section, and designated C. parvipapillatus (C.K.Tseng) Garbary & J.T.Harper as the type species of this section. Several years later, Nam (2006) proposed the genus Palisada following a morphological cladistics analysis that resolved two paraphyletic clades of Chondrophycus species. Chondrophycus parvipapillatus fell into the clade that did not include C. cartilagineus, the generitype, and thus was transferred to Palisada. Interestingly, our specimens from Bermuda exhibit the three traits Nam (2006) used to separate C. parvipapillatus within the genus Chondrophycus: right-angle arrangement of tetrasporangia, possession of secondary pit connections and apical cortical cell projections. Secondary pit connections are noted to be sporadic in P. parvipapillata, but are frequent and conspicuous in the Chondrophycus from Bermuda, and in the former species, cortical cell projections are present throughout the thallus, whereas these are only occasionally seen in the latter, strictly at branch apices. In our analyses, COI-5P sequence data from a Hawaiian specimen of P. parvipapillata nests the species within the Palisada clade ( Fig. 1), providing molecular evidence for its correct placement in the genus and dismissing it as a possible range extension from Bermuda species, despite some morphological similarities.
Of the species of Chondrophycus exhibiting compressed axes, only C. kangjaewonii lacks sequence data available from GenBank. But, this species is easily distinguished from Bermuda specimens by the parallel arrangement of its tetrasporangia, which are also three or four times larger in diameter than tetrasporangia in Bermuda collections. The comparisons of Chondrophycus species with compressed axes are shown in Table 3. Our novel Bermuda species is distinguished from all of these in having upright axes with a length of 2 cm or smaller  and by the conspicuous secondary pit connections present between cortical cells (Figs 14 and 15), a character that historically has been used to segregate morphological subclades of taxa. The novel species of Chondrophycus from Bermuda represents the only current species of the genus that we know of to possess secondary pit connections. However, this trait is not synapomorphic in other genera in the Laurencia complex, including in the sister genus Osmundea. McIvor et al. (2002) suggest that secondary pit connections may be an ancestral state that was subsequently lost in some lineages or species in the complex, a concept that is not easily rationalized given our poor knowledge of pit connection functionality on the whole. succisus displaying both of these patterns and C. dotyi bearing irregular branches similar to some specimens from Bermuda. Our specimens occasionally exhibit cortical cell projections at branch apices, a trait shared by C. dotyi and C. kangjaewonii, but lacking in C. undulatus and C. succisus.
Notably, all of the compressed species of the genus from the Pacific Ocean are found in intertidal habitats (or shallow subtidal, as in C. kangjaewonii). Our Bermuda specimens have been collected exclusively in subtidal waters off the south shore of the islands, from depths between 13 and 23 m during the winter from December to March, when water temperatures in Bermuda are at their lowest. These findings suggest that this novel species is adapted to cooler water temperatures and lower light levels than its most similar congeners.
The genus Laurenciella was segregated from Laurencia (Cassano et al. 2012b) on the basis of molecular sequence data, which showed it to be a distinct clade despite After a thorough assessment of species in Laurencia reported for the region, we determined that our newly collected western Atlantic specimens are unique and cannot be attributed to an existing epithet in the 'Laurencia-complex' for which molecular data is unavailable. We did consider the historical name Laurencia obtusa var. gracilis (C. Agardh) Kütz. based on the illustrations and Latin description of Kützing (1865), which we interpreted as "delicate branches spreading widely," and of  who describe it as a "delicate, soft and slender form." A request was made to the Lund Herbarium for scanned images of the original Chondria obtusa var. gracilis that C. Agardh reported from the West Indies, but unfortunately we have been unable to obtain this historical material to date. Thus, we believe linking this variety (which we did not observe in our collections from St. Croix in the West Indies) to our specimens would be unwise at present.
A second genetically distinct species of Laurenciella from Bermuda has been exposed in this study, resolving as sister to L. marilzae (1.4% divergence) with moderate support. This entity, listed as 'Laurenciella sp. 2 Bermuda' (Figs 1, 2) for the present, groups closely with an undescribed species of Laurenciella (Cassano et al. 2012b) from Brazil (0.9% divergence). As with Yuzurua sp. 2 Bermuda, we have thus far confirmed a single collection representing this taxon, and have conservatively refrained from describing it as a third species in the genus until additional material is collected from the region. HOLOTYPE (designated here): BDA 0934 , March 13, 2012, Gurnet Rock, mouth of Castle Harbour, 32˚20'22.7"N, 64˚39'44.8"W, 13 m. Holotype DNA barcode: GenBank XXXXXX, COI-5P; GenBank XXXXXX, rbcL PARATYPES: Specimens listed in Table 1 as CWS/ .     indicates less than 50% support. Specimens with data generated for this study appear in bolded text.

Palisada flagellifera Brazil
Laurencia cf. mcdermidiae New Caledonia        Two genetic species that are both morphologically identified as the red alga

Laurencia intricata
Laurencia obtusa are not genetically allied to specimens of that taxon collected near the type locality (England). These have been determined to be L. dendroidea and L.
catarinensis, both with type localities in Brazil. Thus, L. obtusa has been removed from the Bermuda flora. We have presented the first report of the Indo-Pacific species Palisada flagellifera in the islands. We have also described two novel species in the 'Laurencia complex'-Laurenciella namii, a second species in the newly erected monospecific genus Laurenciella, and Chondrophycus planiparvus, which also represents the first report of this genus in Bermuda and in the Atlantic Ocean to be verified with molecular sequence data.
In the red algal family Liagoraceae, we have proposed the addition of three genera to the Bermuda flora-Hommersandiophycus, Trichogloeopsis and sibling endemic species are discovered from half a world away (see Appendix B). This is certainly a trend that warrants further exploration.
What little we have learned from this small chain of islands in Bermuda could have broad implications for the Caribbean flora to the south, and also suggests greater levels of endemism in Bermuda than previously thought. While many Caribbean algal taxa are considered to be pantropical, molecular data may reveal certain groups that appear to be more limited in scope than currently accepted, warranting future biodiversity studies in this region. Additionally, because we suspect that such a large number of marine plants are yet to be described in the western Atlantic, it is not unreasonable to predict that a significant number of novel species will continue to be exposed both in Bermuda and the tropical seas of the Americas. Our results from examining the Liagoraceae in Bermuda, as well as preliminary data suggesting additional cryptic diversity in the Caribbean, demonstrates the need for extensive collections in the order Nemaliales throughout the western Atlantic. In the 'Laurencia complex', more work has been done by groups in Mexico and Brazil, but the entire Antillean chain needs to be examined further to identify the northern and southern distributional limits of the Caribbean flora. Extensive sampling of the Greater and Lesser Antilles, areas poorly studied for molecular systematics, would be instructive in piecing together the biogeographical implications of this work for the warm temperate western Atlantic.
Though practice of MAAT and the type method should be routine for taxonomists, it is unfortunately not always so simple. The ability to make these important comparisons depends on either (1) genetic sequence data for organisms from the type locality being available in a public database, such as GenBank, or (2) having connections with a network of colleagues around the globe who have the ability to access and collect material from type localities so that data can be produced if it is not currently available in databases. Often these can be problematic, because (1) the open-source databases are missing a considerable number of the known species, as well as inconsistency in the molecular markers used to make comparisons between lineages, and (2) phycologists may not have colleagues in proximity to a particular type locality, or if they do, the specimens may be rare or difficult to access, or the person may not have the expertise necessary to identify and collect the proper material. Hopefully, these challenges will become more conquerable as the practice of analyzing and publishing molecular sequence data in phycological studies becomes the rule rather than the exception. Accordingly, we should look toward improved procedures for making data available and relatively standardized, to facilitate meaningful use by the phycological community. The establishment of digital museums where the morphology, geography, ecology and genetic information associated with biological collections can be accessed, analyzed and synthesized by researchers is a compelling prospect with potentially far-reaching outcomes.
The precision of biotic and genetic inventories may be especially important in the face of climate change, since inaccurate baseline surveys will mask species loss.
The global climate has increased in temperature approximately 0.74 ± 0.18°C over the past century, with further increase of 2.0 to 4.5°C predicted in the next century (IPCC 2007). Numerous studies indicate that rising sea temperature is affecting the geographic ranges of various marine animals (Hendriks et al. 2006, Parmesan 2006.
Many of these studies target large, migratory species such as turtles and fish, which may be directly responding to temperature, or to a variety of indirect factors like oceanographic currents or availability of prey. To get a better understanding of the potential impact of rising sea temperature in coastal habitats, we should study the Agardh]. Furthermore, warm water species that are presently scarce in Bermuda may become more competitive, and as sea temperatures increase in the next several decades, could become dominant members of the flora with unpredictable consequences.
The creation of a baseline dataset to compare against future algal invasions or losses due to changing water temperatures may become an essential reference tool for the future monitoring of marine biodiversity in Bermuda, as well as southeastern North America and the Caribbean. Studies that aim to discover new organisms and understand the characteristics and patterns of genetic and/or morphological variation in related organisms, will ultimately allow researchers to measure the condition of individual species and ecosystems over time, informing conservation management and policy when necessary. Emergent scientific research and monitoring in this realm are critical on both local and global scales, as new technologies improve our ability to sustainably manage biodiversity.

Introduction
Molecular-assisted alpha taxonomy (MAAT) has recently emerged as an effective technique due to its ability to conquer the challenges of classifying many organisms prone to simple or convergent morphologies. Here we review its use in marine macroalgae, particularly the red algae (Rhodophyta), a group whose individuals are infamously difficult to identify due to characteristic obstacles such as heteromorphic life cycles, evolutionary convergence, and the influence of environmental factors on phenotype expression. The incorporation of molecular data to the understanding of red algal classification has fundamentally altered the way in which we understand this group. Prior to the advent of gene sequencing, taxonomic placement among red algae often could not be definitive without reproductive structures. Individuals discovered with only vegetative characteristics were often classified by their relatedness to species only at the generic level, an inexact science at best. Likewise, convergent evolution and recent speciation often produce cryptic species -impossible to differentiate at the gross morphological level, but in fact genetically distinct. Molecular analyses can now be used to reveal discrete algal ancestries obscured by seemingly identical appearances as a result of converging on a similar morphology. At times, however, the reverse is true. Many species exhibit great morphological diversity based upon the environmental conditions under which a population of marine algae grows (protected vs. exposed environments, shallow subtidal vs. deep-water habitats) and this diversity is often misinterpreted as multiple species. Thus, two fundamental questions guide molecular-assisted alpha taxonomy: Will the alleged conspecifics hold up under molecular scrutiny (Fig. 1a), or are distinct entities present (Fig. 1b)? Or, conversely, will the data confirm or reject apparent differences between uniquely classified individuals (Fig. 1c)?
Although alpha taxonomy is generally performed in similar ways across algal groups, the molecular element of MAAT studies can vary depending on the taxa in question. In some cases, little is known about the species being examined. This is particularly true in instances where reproductive features are unknown, and in such cases, more 'traditional' molecular phylogenetic methods are often used. These include the use of full, or nearly full, gene or ribosomal RNA (rRNA) sequences and in-depth likelihood and Bayesian analysis. In cases where the goal is examining potentially closely-related species, the approach might be to use an established "DNA barcode", such as the 5' portion of the cytochrome oxidase I gene (COI-5') encoded in the mitochondrion, the internal transcribed spacer (ITS) of the nuclear rRNA or a portion of the plastid 23S ribosomal subunit (UPA). These DNA regions may be used individually, or in combination, to determine whether there is an obvious distinction between intra-and interspecies genetic variation, but this method can be subject to sampling biases [e.g., 1]. Accordingly, the sampling strategy for barcoding studies is geared towards obtaining multiple samples of each possible species, in order to better assess species boundaries at the morphological and molecular level. Whereas none of the above methods will give an absolute identity to an alga in every instance, the molecular results can form a framework in which it is possible to re-assess recognized "taxonomically informative" characters or establish novel ones. In combination, each technique can inform the other to establish a more robust classification scheme.
Over the past decade, two of the authors (CWS, CEL) have been applying MAAT to the marine flora of Bermuda in order to better understand the algal biodiversity of this isolated and small island grouping. The Bermudas are among the northernmost islands in the world to support a tropical marine biota [2,3], including the highest latitude coral reefs in the western Atlantic [4]. Eighty-seven percent of the species living in Bermuda are known from the Caribbean Sea, and despite their small size and distance from the tropics proper, the islands of Bermuda support approximately 30% of the 1442 species known in the tropical and subtropical western Atlantic [5], the bulk of them residing in the Caribbean proper. The "lifeline" of the biota of Bermuda is the swift-flowing Gulf Stream, which brings tropical waters from the Gulf of Mexico and Caribbean via the Florida and Antilles Currents, these originating with the Northern and Southern Equatorial Currents off northern South America [2,4,6]. Because of seasonal water temperature oscillations, the macroalgal assemblage of Bermuda is made up of warm-water tolerant species from the western mid-Atlantic that have persisted since re-colonization during the last ice age [7], coolwater tolerant Caribbean species carried by the current northward and an additional 3% that are endemic to this "postage stamp" in the Atlantic Ocean.
Bermuda is a particularly good example of how the use of MAAT can redefine a marine flora because of its location, the history of taxonomic work on the islands and the manageable number of taxa. As a small and isolated archipelago, dispersal to and from Bermuda is relatively rare, leading to an environment amenable to the evolution of new species through genetic isolation. Additionally, despite the heavy influence of the Caribbean on the marine flora of Bermuda, a major obstacle to understanding seaweed biodiversity surrounding the islands has been the North Atlantic biases of early visiting botanists. All of the early descriptions of algal species from Bermuda were reported by individuals from either New England or Europe [8] and many species were misidentified based on a superficial likeness to North Atlantic algae. Many of Bermuda's algal species, therefore, have had to be painstakingly decoupled from the names of their northern relatives. Investigations based on morphology alone, for reasons outlined above, are often inadequate to convincingly reclassify species that cannot be found reproductive, thus making use of MAAT in this process a critical step towards understanding Bermuda's marine biodiversity.
The first compiled list of species in the marine flora of Bermuda was "The Algae of Bermuda" in 1917 by F.S. Collins and A.B. Hervey [9]. This report included 342 species of marine algae, many with names of recognized eastern North Atlantic species. This report [9] was the last complete flora strictly dedicated to the algae of Bermuda. A year later, M.A. Howe [10] contributed the algal section to N.L. Britton's 1918 Flora of Bermuda, but he only included the more common and more conspicuous algae occurring in the islands and again exhibited a northern bias. In the 1940s, W.R. Taylor first visited Bermuda with his student, A.J. Bernatowicz [11], and he included their data along with previous collectors in his 1960 comprehensive [12].

Marine Algae of the Eastern Tropical and Subtropical Coasts of the Americas
Nearly a decade later, Taylor and Bernatowicz [13] produced an annotated list of the most common shallow water macroscopic seaweeds of the Bermudas. At that time, 50 years after the Collins and Hervey report, 40 new species of algae were added to the Bermuda flora and, within the context of the Caribbean, many Bermudian seaweed names were changed. These works brought the total marine red, green and brown species in the islands to 380.
From the time of Taylor's work until late in the 20 th century, additions to the Bermuda marine flora have been published only sporadically. However, beginning in 1997, more than 30 new species and new records to the islands were added to the flora culminating in a checklist of taxa [8]. Since then, nearly 50 more have been added [14][15][16][17][18][19][20][21], bringing the total flora to 449 species. Of this number, 258 are red algae. Several additional questionable records of European taxa reported in the first half of the 20 th century have yet to be confirmed and are not included in the total. It is important to note that 65 species in the flora have their type localities (origin of the specimen used to define a species) in Bermuda, several of which are presently synonymized with other species described from distant geographic locations. Below we review the significant impacts MAAT has recently had on the understanding of Bermudian and western Atlantic seaweed diversity.

Convergence, Hidden Diversity and Sorting It Out
Classification and taxonomy in red algae has traditionally been based on reproductive structures. In particular, the development of female gametangia, or "carpogonia", both before and after fertilization, and the development of tetrasporangia, when found, are extremely valuable taxonomically. The sporophytic stage in the well-known triphasic red algal life history develop sporangia that meiotically produce four tetraspores. These sporangia divide in one of three distinct division patterns typical for genera, and at times, higher taxonomic units. Characters derived from these stages of the life history are critical in differentiating the majority of red algal species, but many collections have either only male reproductive features or none at all, making morphological convergence of two species at the gross anatomy level a vexing problem.

Convergence
A red algal species recently discovered in Bermuda, Chondracanthus saundersii C.W. Schneid. et C.E. Lane (Gigartinales, Gigartinaceae), was discovered devoid of female gametangia or tetrasporangia [15]. Although male spermatangia were present in some collections, these were not helpful for generic assignment. As a genus, the taxonomy of Chondracanthus Kütz. was first revised morphologically [22], then recently using the large subunit of the chloroplast-encoded RuBisCo gene (rbcL) [23].
After rbcL sequencing of Bermuda specimens from Walsingham Pond and Harrington Sound collected at various times of the year, these collections grouped within, but were not identical to, other Chondracanthus sequences [15]. The collections of C.
saundersii were found to be morphologically identical to specimens historically collected in Bermuda, and throughout much of the Caribbean, known as C. acicularis (Roth) Fredericq. However, at the molecular level the western Atlantic specimens are quite distinct from sequences produced from isolates collected in Europe, the area from which C. acicularis was originally described. Based on the morphological reexamination of material from both sides of the Atlantic, character differences were discovered. The primary morphological distinction between C. acicularis and C.
saundersii is the narrower, flattened lubricous axes and less dense medulla of C.
saundersii, but these characters alone would not have convinced all taxonomists of the distinction between the two species using strictly traditional practices.
Chondracanthus saundersii is also smaller and less copiously branched than another European Chondracanthus once also thought to occur in Bermuda, C. teedei (Mert. ex Roth) Kütz. [15]. Chondracanthus teedei was likewise removed from the flora of Bermuda after morphological observations of historical specimens from the islands also showed them to belong to the newly described C. saundersii. Thus, molecular work in Bermuda became the platform upon which European C. acicularis and C. teedei were questioned or removed as members of the western Atlantic flora.
Collections of C. acicularis from Haiti, Cuba, and Brazil were also annotated as C.
saundersii, thus affecting their floras. The Bermuda focused study [15] did report two western Atlantic records (North Carolina, Florida) that for the present remain as C.
acicularis, but even these locations should be sampled and their genetics compared against European sequences of this species in order to confirm that this species remains as a true member of both the eastern and western Atlantic floras.

Hidden Diversity
In a similar manner, Botryocladia pyriformis (  . Botryocladia bermudana, presently an endemic to Bermuda, recently became the sixth member of the "B.
pyriformis complex," again using MAAT [17] .This species is by far the most common of the three Botryocladia species in the islands [17], and is found at depths ranging from the intertidal to 73 m. But, since its first report in Bermuda by Collins and Hervey [9] and Howe [10] [17]. This. molecular sequencing allowed the discovery of this new Botryocladia with minimal anatomical differences from the species it had previously been assigned to in Bermuda [17].
To make matters worse for traditional taxonomists, herbivores in the natural environment of Botryocladia bermudana beget mostly populations of algal individuals under 1 cm tall, creating the illusion of a second species when compared with 6.5 cm, highly branched individuals residing in a protected environments, such as the reef tank in the Bermuda National Aquarium. The Bermuda Aquarium has only ever housed anything in its tanks that has not been collected locally, making it an occasional sanctuary for native marine algae that end up in tanks lacking herbivores. It appears that the parrot fish which graze B. bermudana in "wild" habitats have been an important recent force shaping the scattered B. bermudana populations, as Collins and Hervey [9] never collected its smaller "cropped" form, instead collecting luxurious large specimens in inshore environments where today they cannot be found [17].
Stunted or cropped growth has also been observed among offshore Botryocladia carposporophytes [17]. Unfortunately, no molecular data exist for B. macaronesica nor was any available for DNA extraction at the time of the study [17]. While there is no single outstanding morphological feature which separates B. flookii from all other Botryocladia species, the genus Irvinea was split from Botryocladia on the basis of molecular genetics alone [28]. Since this split, confidence in the use of morphology to separate Botryocladia species has seriously diminished.
Another red alga that contained cryptic species within its binomial in a variety of locations worldwide, including Bermuda, was Asteromenia peltata (W.R. Taylor) Huisman et A. Millar [28]. This species was first described from Venezuela based solely on vegetative material [30,as Fauchea peltata W.R. Taylor]. Since that original report from the Caribbean, A. peltata has been reported as a cosmopolitan tropical species [29]. Recent molecular work [21], however, has carved up 'A. peltata' into five regional Asteromenia species aligned with the Hymenocladiaceae rather than the Rhodymeniaceae [31]. During this recent study [21], two species of Asteromenia has been reported from and not sequenced in this study are now in need of molecular analysis to discover which species they actually represent, as A. peltata has only been verified using gene sequencing in the western Atlantic Ocean [21]. Once again, molecular work created a domino effect of species in need of additional clarification.
Centroceras clavulatum (C. Agardh) Mont. (Ceramiales, Ceramiaceae) has been a reported cosmopolitan alga in the biogeographical literature [3,32]. This species has been recorded from the Pacific, Atlantic and Indian Oceans and places such as Pacific South America, California, North Carolina and Bermuda to Brazil, southeast Asia, Australia, the South Seas and all of the coasts of Africa, to name a few [33]. Thus, it ranges from cool temperate to tropical waters [3]. This was the case until, once again, MAAT greatly restricted the biogeography of C. clavulatum, in this case to the Pacific Ocean [32]. Molecular phylogenetic trees illustrating the relationships between various Centroceras specimens were produced using small subunit (SSU) and large subunit (LSU) nuclear encoded genes, and ribulose bisphosphate carboxylase (rbcL) chloroplast DNA. Three significant changes were made in the taxonomy of this species described in the 19 th century [34,as Ceramium clavulatum C. Agardh]. First, true Centroceras clavulatum was found to be restricted to the Pacific coasts of the Americas from southern California to Chile, southern Australia and New Zealand. Because C. clavulatum had previously been reported from Bermuda, these collections now required analysis both morphologically and molecularly in order to determine their true identity. Next, this study [32] resurrected three western Atlantic species from the 'C. clavulatum complex' which had previously been retired as synonyms of C. clavulatum: C. gasparinii (Menegh.) Kütz., C.
hyalacanthum Kütz. and C. micracanthum Kütz. [35] . Lastly, Won et al. [32] also added two new species to the genus, one from South Africa and the other from southern Chile.

Morphological Plasticity
Halymenia pseudofloresii Collins et M. Howe is a noteworthy Bermudian red algal species because of its distinctive morphological plasticity. Prior to molecular analysis, a similar species, H. floresii (Clemente) C. Agardh, was also a member of the Bermuda flora [8]. Nuclear LSU ribosomal DNA and a protein-coding elongation factor were sequenced and compared with sequences from other Halymeniales using baeysian and likelihood molecular phylogenetic methods. The cytochrome oxidase subunit I (COI-5P) from the mitochondria and universal plastid amplicon (UPA) were likewise sequenced from five Bermuda isolates including H. pseudofloresii and H.
floresii morphs in order to investigate the possibility of intraspecific divergence [19].
The type locality of H. pseudofloresii is not precise, being described as from "a grotto near Walsingham" [36], clearly within the national park of that name in Bermuda.
There are a number of ponds and pools (or grottos) in the Walsingham area, all of which are connected by a system of underground saltwater caves. Several specimens from this tidally fed system were collected for molecular and morphological analysis.
Molecular analysis showed no more than a single base pair difference between the COI-5P of any two of the isolates, which included distinct morphologies such as narrowly pinnate (H. floresii morph), intermediate, and typical broad fronds (H. pseudofloresii morph). Likewise, no differences were observed in the UPA of any of the Bermuda specimens tested [19]. Thus despite the range of appearance, molecular data from several common markers clearly indicates that they all represent a single, phenotypically plastic species in the islands [19]. Additionally, all of the Bermuda plants were distinct from a sequence obtained from freshly collected material of H.
floresii from near its type locality in Spain, thus bringing into question other reports of this European species from the western Atlantic. Further comparisons with sequences from other members of the Halymeniales in this study [19] suggest that Halymenia is a polyphyletic genus as it currently stands, opening up another avenue for molecular work. Clearly, other species found in Bermuda and elsewhere which are observed as morphological continua between distinct morphologies could benefit from the refined techniques of MAAT [19]. Some may, in fact, represent distinct species with overlapping morphologies, while others such as H. pseudofloresii outlined here represent species with highly variable mophologies.

Remaining Barriers to Sorting it All Out
Clarifying taxonomic issues often requires more than just obtaining samples of the species in question. It is often necessary to examine type specimens and then critical to collect and process DNA samples from the type locality and from potentially related species. For that reason, many species remain in limbo while comparative collections are being made. This process can take years, because some species have only been collected once or exist in remote locations. Additionally, habitats have changed since the time early 20th century collections were made and locating comparative material can be impossible. This is especially a problem because early vegetative algal collections, in particular, had their broader taxonomic placements misjudged as often as they were erroneously classified at the genus or species level. For example, when Taylor [37] described the genus and species Rhododictyon bermudensis W.R. Taylor from Bermuda, he tentatively placed it in the family Dasyaceae (Ceramiales). This new genus and species was later moved to the Ceramiaceae prompted by the discovery of tetrasporangia in offshore collections from North Carolina [38]. The tetrasporangia of R. bermudensis form at the ends of filaments obtruding from the lower, older cells of the blade. Such an arrangement is similar to that of Compsothamnion Nägeli and was used to propose an alliance with the tribe Compsothamnieae [38]. Because taxonomically critical female gametangia have yet to be discovered for this genus, molecular analysis is required to firmly establish the taxonomic position of this monotypic genus in its family and tribe [38].
Yet this species is only found at great depth and in small numbers when found, an example of a microscopic taxon difficult to find enough material for both DNA and morphological study.
Flahaultia tegetiformans W.R. Taylor (Gigartinales, Solieriaceae) provides an example of a red alga that may not have been placed in the correct genus. Despite numerous collections from a variety of locations in Jamaica none were ever collected with gametangia [39]. Taylor described this species as a prominent member of the Jamaican flora from three subtidal sites. Again, Taylor's pressed herbarium collections were made prior to sample preservation for DNA extraction, and no recent workers have collected fresh material for genetic sequencing. The Jamaican vegetative specimens sat upon Taylor's lab bench for years awaiting fertile collections that were never found; eventually, he assigned the new species to the genus Flahaultia Bornet on the basis of its resemblance to the eastern Atlantic F. appendiculata Bornet.
Although a recent report of F. tegetiformans from the Greater and Lesser Antilles has been made [40], no specimens have been sequenced from either the type locality nor these recently collected sites. Such an analysis is necessary to ensure that the Antillean specimens are indeed the same as those collections from the type locality in Jamaica, and also whether the species is properly assigned to Flahaultia. Thus, the taxonomic placement of F. tegetiformans remains uncertain.
Saunders appears in habit and vegetative morphology very similar to Flahaultia tegetiformans, sequences of its SSU nuclear DNA from Bermuda were used to firmly place it in the Sebdeniaceae (Sebdeniales), rather than the Solieriaceae. This classification was affirmed by the remarkable similarity in female sexual structures and post-fertilization stages observed between C. walsinghamii and Sebdenia (J. Agardh) Berthold when female material was found in a collection discovered after molecular sequencing was completed [18].
Another recent discovery in Bermuda was the endemic Griffithsia aestivana C.W. Schneid. et C.E. Lane [16]. It was identified without the benefit of gametophytes or gene sequencing, but was placed in Griffithsia C. Agardh owing to its obvious morphological resemblance to G. capitata Børgesen from the Canary Islands and other like members of the genus. The new species was based upon the formation of tetrasporangia located on whorled stalk cells at the distal ends of axial cells, among other things. Whorls occur around the two penultimate cells at the tips in the similar G. capitata; whorls occur on long cylindrical cells from mid-axis to a few cells below an apex in G. aestivana. Although G. aestivana seems to have unique characteristics, molecular analysis of G. aestivana is still necessary to ensure that it is a unique species, and the taxonomy will not change until that is completed [16].
Despite the fact that MAAT is becoming a standard procedure in floristics and phylogenetics, not all new species are being preserved for DNA extraction. Some are simply so small and of unknown identification in the field, that they remain to be determined in the lab long after the collections are made. Woelkerlingia sterreri C.W.
Schneid. et M.J. Wynne was recently described from Bermuda [20]. This species was collected at 10 m, extremely well established on the unusual habitat created by a freemoving discarded linen tablecloth. The presence of distinctive 2-celled fertile female filaments and other characters of the procarp were critical in placing the specimens in the genus Woelkerlingia Alongi, Comaci et G. Furnari. However, no molecular work has been done within this genus to corroborate its taxonomic assignment; therefore, it is questionable whether or not all the genera of small fuzzy reds in the Wrangeliaceae are supported genetically, including Woelkerlingia [20].

Conclusions
The barriers to understanding algal biodiversity are problematic on a global scale for reasons covered here. Even relatively heavily studied genera in marine floras that were thought to be well understood using morphology and anatomy, have been revised based on recent MAAT studies [41][42][43], suggesting that substantial taxonomic revisions are likely over the next few decades. Although the MAAT method will almost certainly be superseded by novel techniques as methodologies advance, the combination of molecular data and microscopical observations has proven to be a robust approach to solving many long-standing taxonomic controversies among algal groups, particularly within red algae. Ultimately, the extent to which MAAT studies expand will likely depend on the stabilization of markers for various groups of the Tree of Life and continued training of students in alpha taxonomy -a skill that is rapidly declining in prevalence as descriptive morphological taxonomic work has fallen out of favor. The strengths of MAAT studies are, like any type of research, entirely dependent on the input data. DNA barcoding has been lauded as a major step forward in species identification, but several studies [1,44] have shown that sampling biases, either geographic or taxonomic, can result in incorrect species delimitations.
What is acceptable in one lineage of organisms, however, may not work in another.
Ironically, the very diversity of Eukaryotes confounds our attempts to apply a universal marker to understand that diversity.
However, the short-term importance of understanding global biodiversity in an era when it is being depleted means that methods to do so must be identified,

INTRODUCTION
The isolated western Atlantic islands of Bermuda (32˚18'N, 64˚46'W) have proved to be not only rich in macroalgal biodiversity, but also host to an invaluable reserve of genetic information for taxonomists. More than 70 species of benthic marine algae, as well as 15 subspecific taxa, have their type localities in Bermuda, these representing nearly 20% of the local flora and 5% of the approximate 1400 tropical and subtropical marine algae in the western Atlantic (Wynne 2011  .
Shortly after the genus Crassitegula was proposed for a single species from Bermuda ), a second species was described from almost half a world away at Lord Howe Island, eastern Australia (Kraft & Saunders 2011
A single collection of this species was made in a species-rich habitat off the north shore of Bermuda. It conformed in all macroscopic and anatomical characteristics to Anadyomene lacerata (Littler & Littler 1991  , also has elongate "vein" cells along the outer margins (Littler & Littler 1991). This latter species, however, has a pattern of randomly organized, oval-shaped interstitial cells between the veins (Littler & Littler 1991), in contrast to those in A. lacerata which are elongated and parallel, basically perpendicular to the vein cells, and separated by rows of shorter interstitial cells (Figs 2, 3) (Littler & Littler 1991).
This collection from Bermuda represents a new northernmost range extension for this offshore subtidal species, previously known from 15-60 m depths in the Gulf of Mexico, the Caribbean Sea, and off the coasts of Brazil and Venezuela (Littler & Littler 1991).

Bryopsidales, Codiaceae
Codium carolinianum Searles 1972, p. 19, figs 1a, 2. Not previously known from Bermuda, recent collections from a seemingly persistent offshore mat of this prostrate species are a good morphological match to specimens from Onslow Bay, North Carolina, USA (the type locality). We obtained and sequenced the rbcL portion of the chloroplast genome from two offshore North Carolina specimens for comparison ( were the basis of the report of G. curtissiae from Bermuda  Louisiana, USA, as identified by  and , and posted to GenBank as AY049322. This report effectively removes G. curtissiae from the flora of Bermuda and represents a new northern limit of distribution for G.

Nemastomatales, Nemastomataceae
Crassitegula laciniata was easily distinguished morphologically as it was more branched and dissected than both the generitype and its remarkably similar southern hemisphere congener, C. imitans (Kraft & Saunders 2011). The latter two species have suborbicular proliferations at the margins, this being the form taken as they spread over rock surfaces whether the surface is vertical or horizontal. Conversely, C. laciniata in early development was seen as small transversely ovate to angular blades that produce occasional minute marginal dentations, most of these distally, that develop long, narrow finger-like branches (Figs 13,15). As the plant matures, these narrow, acute-tipped branches broaden to become strap-shaped and often dichotomous (Figs 14,15). Thus, mature thalli totally differed in habit from the juvenile blades that initiated them (Fig. 14).
We made several collections of this new species in spring and summer, mostly off the north shore of Bermuda, and although each consisted of numerous individuals, no tetrasporophytes or gametophytes have been encountered such as have been detailed for the other two Crassitegula species by  and Kraft & Saunders (2011). Therefore, molecular data have been necessary to characterize the generic placement of the third species, which, as already stated, conformed to the genus Crassitegula in its anatomical features (Figs 16,17).
As pointed out in Kraft & Saunders (2011), the Sebdeniaceae is now much richer in genera and species than it was when Kylin (1932) created the family for the single genus Sebdenia. Its monogeneric status persisted until  added the genus Crassitegula, which was soon followed by the addition of Lesleigha by Kraft & Saunders (2011). Significant diversity in the family has been discovered from the Indo-Pacific region, from Australia to Hawaii, where Lesleigha and its three species were recently described (Kraft & Saunders 2011). In that same paper, C.
imitans was described from Lord Howe Is., off the eastern coast of mainland

INTRODUCTION
In the early 20 th century when the first modern compilation of algal species for Bermuda was completed by , only a single Centroceras species was recorded, at that time regarded as Ceramium clavulatum C. Agardh. This species was thought to be the "commonest and most variable Bermudian representative of the Ceramiaceae" (Howe 1918). Howe (1918)

recognized that
Centroceras clavulatum (C. Agardh) Mont. was distinguished from Ceramium by its nodal spines, and that their "length, shape, and abundance" varied greatly among individuals found in the many different habitats of the islands. Despite this, at the time  completed his comprehensive flora of the warm waters of the western Atlantic, only a single species of Centroceras was again reported. Since then and prior to the advent of molecular sequencing, only two other species had been added to the western Atlantic flora, C. internitens Gallagher et Humm (Gallagher and Humm 1983) and C. minutum Yamada (Littler et al. 2008), both small epiphytes that could not be confused with the larger C. clavulatum.
Until recently, Centroceras clavulatum was considered a poster child for cosmopolitan marine algae (Lüning 1990, van den Hoek andBreeman 1990). It had been recorded from the Atlantic, Pacific and Indian Oceans , and from warm temperate to tropical waters (Lüning 1990 clavulatum was until now one of just two species reported from Bermuda, island collections were in need of both molecular and morphological assessment to determine their taxonomic identities, the focus of the present study. interface (Silvestro and Michalak 2012) with 1000 bootstrap replicates. Bayesian posterior probabilities were generated in MrBayes, with the same parameters as in the COI-5P analysis. Stationarity was attained after the first 250,000 generations (burnin=2500 trees). The COI-5P (Fig. 1) and rbcL (Fig. 2)

TAXONOMIC RESULTS AND DISCUSSION
Based on our many Bermuda DNA samples and available sequences from GenBank (Table 1), COI-5P barcode analyses ( Fig. 1) and rbcL phylogenies were generated (Fig. 2). As shown in these trees, the three species  resurrected for the western Atlantic Ocean by aligning type specimens with their recent genetic material, Centroceras gasparrinii, C. hyalacanthum and C. micracanthum, are found in the Bermuda flora. In addition, two novel species were delineated in our molecular analyses. During the 20 th century, all of these five species would have been identified as C. clavulatum in the islands (Schneider 2003, p. 301).
Along with the new genetic information, we follow Won et al. (2009, fig. 14) in using nodal cortical units and the origin, number and shape of cortical cells, glands and spines produced from cortical initials as distinguishing features among the species.
The descriptions presented for each species below reflect genetically determined specimens from Bermuda, Florida and the U.S. Virgin Islands. tetrasporangia formed in whorls at the nodes (Fig. 3D), one produced from each periaxial cell, spherical to subspherical, 26-55 µm diam. and 38-75 µm long, subtended by 0-3 involucral filaments; gametangia not seen.

Centroceras arcii
in a separate clade according to COI-5P and rbcL analyses. Sequence divergence between the new species and C. micracanthum is ~1% for rbcL and ~9% for COI-5P, the latter clearly demonstrating genetic separation.
The most obvious morphological feature of Centroceras arcii is its divaricate, reflexive branching (Fig. 3B) giving the plants an arching overall habit, a feature found in all mature collections of the species thus far. Unfortunately, C. micracanthum demonstrates at least some reflexive branching in ca. 75% of our collections and when this is the case, this species is difficult to distinguish from C. arcii. Won et al. (2009, fig. 7b) demonstrated reflexive branching in C. micracanthum. Thus, C. arcii and C.
micracanthum are cryptic in their anatomies (Table 2). At the extremes of ranges for certain characters, e.g., spine length, number of cells in descending cortical files and median sizes of tier cells, there are differences between the two species, but for the most part, they demonstrate overlapping ranges for cell dimensions and numbers (Table 2). At the extreme, cells composing basipetal cortical files are more numerous in C. arcii than those in C. micracanthum. Invariably, C. arcii has one or two branches formed in the notches of distal dichotomies, and such branches are uncommon in C.
micracanthum. In the end, the two species are nearly indistinguishable and, as noted above, require gene sequencing to differentiate them with certainty.
As is also true for Centroceras micracanthum, in all but the first periaxial cell cut off in Centroceras arcii, three cortical initials form, two acropetally and one basipetally (Fig. 8A). The first periaxial cell cuts off three cortical initials acropetally and one cortical initial basipetally for a total of four cortical initials. In all cortical initial groups, the first cortical initial cuts off one spine and a single elongate cortical cell, one gland cell and a single elongate cortical cell, or two elongate cortical cells.
The production of cortical cells from acropetal cortical initials in C. arcii is the same as that found in C. micracanthum and C. illaqueans, all producing only two cells from their first cortical initials. In C. arcii, the largest spines at nodes are longer than those of any other members of the 'C. clavulatum complex' in Bermuda. ( Table 1).

Remarks.
Multiple Bermuda isolates clustered with Florida, USA and Caribbean sequences of Centroceras hyalacanthum in our COI-5P and rbcL trees (Figs 1, 2). In Centroceras hyalacanthum, the first periaxial cell cuts off three cortical initials acropetally and one cortical initial basipetally for a total of four cortical initials. On the rest of the periaxial cells, only three cortical initials form, two acropetally and one basipetally (Fig. 8C). In all cortical initial groups, the first cortical initial cuts off one spine and two elongate cortical cells, one gland cell and two elongate cortical cells, or three elongate cortical cells. This production of cortical cells from acropetal cortical initials is unique among all of the species thus far isolated from Bermuda (Figs 5B, 8C), the rest producing only two cells from their first cortical initials. In addition, descending cortical cell files are made of elongate rectangular cells (Fig. 5D). This developmental pattern of the cortex initiated at the nodes is in perfect accord with that demonstrated by  for the species including their observations on the lectotype specimen they designated. Amongst all of the species in the genus found in Bermuda, C. hyalacanthum has the narrowest axes, 110-170 µm in diam. (Table 2), the most delicate, from Florida and the Caribbean, being even narrower than those reported by , at 140-165 µm in diam.
In the same paper in which Kützing (1842) wrote the protologue for Centroceras hyalacanthum, he also described an additional Caribbean species, C. oxyacanthum Kütz. (type locality = Cuba), and later described C. brachyacanthum Kütz. (type locality = Antilles, West Indies) (Kützing, 1863). After an examination of types of all three species,  considered the latter two species to be heterotypic synonyms of C. hyalacanthum. Distribution. Endemic to Bermuda and currently known only from the type locality.

Remarks.
Centroceras illaqueans is genetically distinct from all the other sequences in our analyses, and is grouped with C. hyalacanthum in our trees (Figs 1,2). The new species is distinguished from other Centroceras species in Bermuda by the length of its terminal elongated acropetal cells which are mostly equal to or longer than spines at the nodes ( Fig. 6D & E, Table 2), and the alignment of its basipetal filaments which typically line up with other descending files in tiers of short isodiametric cells (Fig. 6D & E). In addition, the new species differs from C.
hyalacanthum by its fewer number of spines at the nodes, spine length and acropetal cell length (Table 2). A recently described species from Natal, South Africa, C.
hommersandii, shares its small size with C. illaqueans, but has axes that are narrower in diameter than the Bermuda species and produces only a single elongated acropetal cortical cell in each cortical cell unit, being produced from the second cortical initial ). The first cortical initial of C. hommersandii produces a spine or gland cell and an ovoid acropetal cortical cell , unlike the elongated cell produced by C. illaqueans.
Despite the numerous collections of Centroceras sequenced from Bermuda, C.
illaqueans was found only once. In this extensive population, the tips of erect axes emerged approximately one half centimeter out of sandy sediment on the bottom of a large, shallow tidal pool on the south shore. Repeated visits to the type locality have not uncovered this species again. periaxial cells cutting off two cortical initials acropetally and one basipetally (Fig. 8E); the first cortical initial cutting off one spine and one elongate cortical cell, one gland cell and one elongate cortical cell, or two elongate cortical cells, the second cortical initial cutting off one elongate acropetal cortical cell and one basipetal cortical filament, the third cortical initial cutting off one basipetal filament; basipetal filaments made up of mostly staggered cortical cell files of 5-17 cells, the cells elongate rectangular in surface view, highly variable in size, from 2-12 µm diam. and 14-116 µm long, these completely corticating axial cells from node to node (Fig. 7D), at times these cells lined up in tiers; gland cells ovoid; gametangia and tetrasporangia not seen in Bermuda collections. This species is best distinguished from C. arcii by its COI-5P barcode sequence (see Table 1 for GenBank numbers).  Table 1).

Remarks.
In our rbcL analysis, sequences isolated from Bermuda specimens grouped with Centroceras micracanthum from Florida and the Caribbean (Fig. 2), and our morphological characters matched those described from the same areas ). In the same work with the protologue of C. micracanthum, Kützing (1842) also described C. leptacanthum Kütz. (type locality = Genoa, Italy), C. cryptacanthum Kütz. (type locality = Antilles, West Indies) and C. macracanthum Kütz. (type locality = Brazil), all of which are now considered heterotypic synonyms of C. micracanthum after an examination of type material ). Given its widespread distribution in the western Atlantic , it is not surprising that C.
micracanthum is found in Bermuda with many other macroalgal species also known in the Caribbean. Based upon morphological characteristics, Florida specimens of C.
clavulatum distributed as no. 1347 in the Collins et al. (1906) exsiccatae, Phycotheca Boreali-Americana, are better identified as C. micracanthum. The label noted that these Key West specimens correspond to C. leptacanthum, a species that Won et al.
As discussed above, Centroceras micracanthum is most difficult to distinguish morphologically from C. arcii in Bermuda. Its habit of reflexive branching in about 75% of specimens and overlap of anatomical characteristics make it truly cryptic with C. arcii. Therefore, this species is best identified by comparing genetic sequences with those available in the public domain.

CONCLUSION
The use of molecular-assisted alpha taxonomy has allowed for the resolution of five species in the 'Centroceras clavulatum complex' in Bermuda. Won et al. (2009, fig. 14) used stylized diagrams of nodal cortical units for each of the species they covered in their seminal study of this complex. They depicted the origin, number and shape of cortical cells, glands and spines produced from cortical initials as distinguishing features among the species, and we have done the same for nodal cortical initials bearing spines (Fig. 8), and the developmental patterns of our two new species can be compared in this way. Three of the five Bermudian species are recorded in their most northerly locations in the western Atlantic, and the remaining two, C. arcii and C.
illaqueans, are newly described from the islands. Studies in other parts of the world's warm temperate to tropical seas will undoubtedly extend or shrink the distributional ranges of some Centroceras species, and thousands of herbarium specimens in the genus worldwide will need further study and annotation. Only then will we have a more accurate picture of the biogeography of the many species in the genus.