The Hawaiian Rhodophyta Biodiversity Survey yielded only a few new taxonomic records for the Hawaiian Islands, but many new distributional records. The flora has been remarkably well-studied from a morphological and anatomical perspective, most notably by Abbott [1
], and many of her designations are supported by our molecular investigations. However, many taxa have also been revealed to be in need of in-depth taxonomic study based on our molecular survey data due to instances of suspected cryptic or incipient speciation (e.g. Amansia glomerata
and Spyridia filamentosa
; see results for full list), or conspecificity (e.g. Gracilaria eppihippisora
and G. salicornia
). As such, our survey data provide a unique opportunity to flag Hawaiian red algal taxa for further study. Although identification of these taxonomic issues, rather than full resolution, was the goal of the Hawaiian Rhodophyta Biodiversity Survey, several of these flagged taxa have been investigated in detail under the auspices of the project (e.g. [8
]), and we hope that other researchers will be able to make use of the data for similar purposes, for it is detailed taxonomic studies that will ultimately lead to a floristic revision of the Hawaiian red algae. Also, additional taxonomic expertise for some of the lineages of red algae investigated during the project (in particular, for many of the Ceramiales), will be needed for taxonomic resolution. However, despite adding relatively few new taxa to the species list, we have substantially updated the reported range of many taxa, with 196 new island-level distributional records included from our sampling records.
The vast majority of accessions analyzed were from marine and brackish locations (1,928, or 99%), while 16 were from freshwater habitats and two were from terrestrial locations. This breakdown is close to the accepted consensus that, worldwide, approximately 97% of red algal species are marine [46
The three molecular markers chosen for diversity assessment of our red algal collections (nuclear LSU, mitochondrial COI and plastid UPA) differ markedly in both their degree of conservation and ease of amplification/sequencing, based on the protocols that were applied to our survey samples. Since the Hawaiian Rhodophyta Biodiversity Survey encompassed an evolutionarily diverse set of samples, and we aimed to generate a large number of sequences (i.e. >2,000), we were constrained by the need for minimal variations of protocol for each marker (including primer design, PCR cocktail components/concentrations and amplification conditions). The LSU and UPA markers were sequenced without modification (i.e. according to the primers and conditions described in [27
] and [29
] respectively), and these first two markers yielded the highest and second highest overall number of sequences, respectively. In contrast, the COI marker required the design of an additional reverse primer, and yielded the lowest number of sequences. Ease of data acquisition and performance characteristics of each of the three markers are further examined and discussed in [47
], as well as below.
All clean sequences (those with clearly readable chromatograms that could be assembled with confidence) were retained in HADB, regardless of whether or not they corresponded to the taxon under study; contaminant sequences (e.g. from epiphytes) were re-labelled as such and given new accession numbers. All three markers yielded some contaminant sequences, but the UPA marker yielded the highest number (30), followed by LSU (15) and COI (7).
The LSU marker was the simplest for which to reliably obtain clean sequence data; only a single pair of primers was needed, and most samples were successfully sequenced on the first attempt. The disadvantage of the LSU marker, however, was a lack of reliable species-level resolution (e.g. Figure , Nemaliales LSU sequence data show no divergence between some members of the genera Akalaphycus
). A recent investigation of the characteristics and phylogenetic "performance" of the LSU, UPA and COI markers based on this same data set confirmed that the LSU marker is the most conserved of the three, that it was the easiest for which to obtain sequence data, and illustrated that florideophyte red algal sequence data for the marker became saturated at F84 distances of approximately 0.30 (although there was variation when individual orders were examined) [47
]. The LSU marker was included in the survey to provide representation from the nuclear genome (both the UPA and COI are organellar markers). However, given that the level of resolution for the LSU marker differs so strongly from the other two markers (Figure ), it would be difficult to recommend its inclusion in future biodiversity surveys.
The UPA marker, which was first described in 2006 [28
] and tested for a selection of algal lineages in 2007 [29
], can be amplified and sequenced for most cyanobacterial and plastid-containing algal lineages, with only a few genus-level exceptions noted to date (e.g. Cladophora
). This universality is exceptional among algal primers, and lends great potential for the use of this marker in assessment of environmental samples that contain multiple algal lineages [48
]. For this reason, we believe it is important to continue to generate framework data for this marker for as many algal representatives as possible. Recent work has shown that the UPA marker falls between in the COI and LSU markers in terms of ease of data acquisition, that it is quite conserved (although slightly less so than LSU), and that saturation is not an issue for any group of the florideophyte red algae based on UPA sequences [47
The COI marker was somewhat more difficult to amplify and sequence across red algal lineages, and two combinations of primers were used to obtain the 639 sequences reported here (although new primers have been designed recently that are generally more successful; G.W. Saunders, pers. comm.). Others have reported requiring up to seven primer combinations to obtain COI sequences for samples within a single family of red algae [49
]. However, the COI marker is widely recognized as the DNA barcode for red algae [9
], and as such, provides a useful sequence data set due to the immediate comparative power of COI sequences against the 11,925 red algal DNA barcode sequences in the Barcode of Life Data Systems (BOLD) v.2.5 database (accessed 05 November 2010 [50
]). This trade-off between universal amplification and resolution has been noted in the past [42
]. We recently reported that these COI florideophyte red algal sequence data were the most difficult of the three markers to obtain (i.e. low ease of acquisition), but that the COI marker is the least conserved (and is especially variable at the third codon position), and that saturation of the entire data set occurs at F84 distances of approximately 0.11 (although this number varies substantially depending on which order is being examined) [47
Maximum likelihood analysis of the concatenated LSU+UPA+COI sequences yielded a phylogenetic tree that included support for most of the florideophyte red algal orders and families included in the data set (i.e. 3/17 orders and 4/37 families were not recovered as monophyletic; Figure ). A number of explanations may be responsible for the lack of full resolution in the phylogeny, including insufficient sequence data [47
], insufficient taxonomic representation, or unresolved identifications of some specimens. Verbruggen et al
], in their recent research on the data requirements for resolving the red algal tree of life, indicated that up to 2.8 × 105
nucleotides of sequence data may be required to resolve some regions of the phylogeny, which is orders of magnitude more data than that provided by our concatenated three-marker data set. We [47
] preliminarily evaluated the phylogenetic performance of the three markers with the same data set by assessing changes in neighbour-joining bootstrap support with the addition of marker data (progressing from UPA, to UPA+LSU, to UPA+LSU+COI), and also concluded that the fully concatenated phylogeny was supported largely at the family and ordinal level, with higher level relationships not resolved. Nonetheless, the sequences of the three markers will make a valuable addition to the red algal species assignment framework of data, and in concatenated form are useful for examining phylogenetic relationships at lower taxonomic levels.
Considering the taxa studied using statistical parsimony analysis (Figure ), the debate arises as to whether observed breaks in networks represent strong phylogeographic signal or species boundaries. It has been suggested that the 95% connection limit under statistical parsimony, which is commonly used in phylogeographic studies, is a useful criterion for indicating biological species boundaries [52
]. Applying this criterion to our analyses suggests that cryptic species may be common in the Hawaiian red algal flora. Only one of the four taxa investigated formed a single haplotype network at the 95% connection limit (Dasya iridescens
), while the remaining taxa formed two (Asparagopsis taxiformis
and Dichotomaria marginata
) or three networks (Amansia glomerata
). If this pattern holds true for a number of other Hawaiian red algal taxa (which will only be determined as more taxa are studied in depth), then there may be many more species than currently recognized in the flora. Sorting out phylogeographic signal from species boundaries requires an understanding of the limits of phenotypic plasticity for each case. Most of our identifications were made, at least preliminarily, based on morphology and anatomy. We largely relied on the detailed descriptions of Abbott [1
] but supplemented these with more recent or taxon-specific studies that were relevant to each taxon (a list of this additional literature can be found in [44
]). Nonetheless, lately there have been a number of cases in the phycological literature where cryptic species complexes have been revealed through molecular data, and in a number of these instances other characters (morphological, anatomical, life history, physiological, etc.) have later been identified that allowed recognition of the species on grounds other than molecular data, e.g. [49
]. More detailed phylogeographic, morphological/anatomical and life history analyses of suspected species complexes will be necessary for confirmation, but the DNA sequences from the present survey provide a valuable starting point for identification of these phenomena, and pave the way for taxonomic revision of the floristic checklist.
Interestingly, not all common Hawaiian Rhodophyta species have notable intraspecific sequence variation. Eight taxa representing three different orders (Table ) were reported to have little to no sequence divergence for the three markers sequenced. Although these taxa require stronger sampling for confirmation and resolution of these trends, one important finding from the survey is that phylogeographic patterns are likely to vary, even for taxa considered to be native to the Hawaiian Islands, and that histories of individual taxa need to be studied (e.g. time since colonization or speciation, colonization order of the islands, interisland dispersal vectors, introduction events from other regions) to obtain a fuller understanding of intraspecific molecular patterns. It does not seem to be the case that increased within-taxon sampling will always reveal obvious phylogeographic patterns. This difference could be due to many factors, including the length of time since colonization of the species (or speciation, in the case of endemic taxa), ease of dispersal, dispersal vectors and type of life history (i.e. whether there is an easily spread life history stage, such as a floating, filamentous form), habitat and physicochemical parameter tolerances.
Ekrem et al. [54
] emphasized that DNA barcode data will only ever be as useful as the reference library of sequences against which comparisons are made. This is a compelling argument for support of biodiversity surveys of entire floras, such as this one, that attempt to account for the diversity of a region with both morphological and molecular sequence characterizations, and that make these data available to others in easily-accessible formats for comparisons. Zhang et al. [55
], however, rightly point out that sample sizes for most DNA barcode studies are insufficient to assess genetic diversity of a species, and the level of intraspecific sampling for our project is no exception. Given the limited time and funds for the survey, we aimed to strike a balance between representation of as many taxa as possible and sufficient coverage of common taxa to assess intraspecific diversity. We emphasize that the patterns discussed here are largely preliminary, and that a more thorough examination is warranted to reveal the full story.
The value of the Hawaiian Rhodophyta Biodiversity Survey is multifaceted: through this project we have generated baseline molecular diversity data for much of the flora, revealed taxa requiring further systematic investigation (both in terms of possible cryptic species that have not been previously recognized, and morphologically/molecularly similar taxa that may be synonymous), begun to elucidate phylogeographic patterns for Hawaiian red algae, revised the species checklist for Hawaiian Rhodophyta, and reported a number of new records (both taxonomic and distributional). All of these data (our own collections plus those from BISH), as well as photographs and micrographs, specimen details and collecting site information, are available via
the publicly accessible Hawaiian Algal Database [56
], and the data have also been shared with the Global Biodiversity Information Facility [57
] and GenBank [58
]. Voucher specimens and the DNA extract library are available to other researchers, enhancing the long-term value of the survey collections. The value of the survey will also increase as other studies make use of the data and our initial observations are developed into additional projects.