Molecular detection and species identification of Alexandrium (Dinophyceae) causing harmful algal blooms along the Chilean coastline

Misleading morphological observations assigned Alexandrium catenella as local dinoflagellate responsible for HABs in Southern Chilean coasts. Our work based on molecular methods found that local Alexandrium belongs to group I of the tamarensis complex composed mainly of A. tamarense.


Background and aims
On the basis of morphological evidence, the species involved in South American Pacific coast harmful algal blooms (HABs) has been traditionally recognized as Alexandrium catenella (Dinophyceae). However, these observations have not been confirmed using evidence based on genomic sequence variability. Our principal objective was to accurately determine the species of Alexandrium involved in local HABs in order to implement a real-time polymerase chain reaction (PCR) assay for its rapid and easy detection on filter-feeding shellfish, such as mussels.

Methodology
For species-specific determination, the intergenic spacer 1 (ITS1), 5.8S subunit, ITS2 and the hypervariable genomic regions D1 -D5 of the large ribosomal subunit of local strains were sequenced and compared with two data sets of other Alexandrium sequences. Speciesspecific primers were used to amplify signature sequences within the genomic DNA of the studied species by conventional and real-time PCR.

Principal results
Phylogenetic analysis determined that the Chilean strain falls into Group I of the tamarensis complex. Our results support the allocation of the Chilean Alexandrium species as a toxic Alexandrium tamarense rather than A. catenella, as currently defined. Once local species were determined to belong to Group I of the tamarensis complex, a highly sensitive and accurate real-time PCR procedure was developed to detect dinoflagellate presence in Mytilus spp.
(Bivalvia) samples after being fed (challenged) in vitro with the Chilean Alexandrium strain.
The results show that real-time PCR is useful to detect Alexandrium intake in filter-feeding molluscs. † Both authors contributed equally.

Introduction
Harmful algal blooms (HABs) occur throughout the world and are known for their negative economic and sanitary impacts (Anderson 2009). Of particular concern are the paralytic shellfish toxins produced mainly by bloomforming dinoflagellates in the genus Alexandrium. Over the past few decades, Alexandrium blooms have extended, covering new territories. In this sense, expansion of dinoflagellate species could be explained by ocean currents, human-induced mechanisms such as water from ballasts and global warming, climate adaptation and colonization of newly generated niches (Anderson 1989). On the other hand, it is also possible that the increasing number of blooms reported today is the result of a worldwide effort to implement new techniques to detect and prevent their negative effects (Anderson 2007). Blooms of different Alexandrium species have been reported from Japan (Kodama 2010), northwestern Mediterranean Sea (Vila et al. 2001), Australia (Hallegraeff 1998), Caribbean Sea, off the Venezuelan coast (Halstead and Schantz 1984), Brazil (Persich et al. 2006), along the American Pacific coasts, from Alaska to the Strait of Magellan (Suá rez et al. 2002;Glibert et al. 2005;Herná ndez et al. 2005), and from north Atlantic coasts from the Gulf of St Lawrence to North Carolina . Compared with the numerous studies describing Australian, North American and Japanese Alexandrium ribotypes and morphotypes, relatively few works describe their South American counterparts from the South Pacific (Lilly 2003). Historically, Alexandrium species were described based on microscopic observations of morphological features including plate patterns, cell size and shape, and secondary characteristics such as chain formation. Unfortunately, these morphological traits have often proven insufficient for identifying species, leading to confusion concerning the distribution, ecology and toxicity within this genus (Lilly et al. 2005). When morphological features are questionable for taxonomic identification, they must be combined with molecular data for accurate species definition (Hansen et al. 2003).
Sequence variation analyses have been accepted to be a valid methodology for an accurate species description. This is even clearer in the Alexandrium genus, which has been partially reclassified on the basis of molecular genetic data, as the taxonomic value of only morphological characters proved to be insufficient for the tamarensis complex (Leaw et al. 2005). A good example is the taxonomic trait 'presence or absence of the ventral pore', used to discriminate between A. affine and Alexandrium tamarense. This trait would not be deemed useful for species identification considering that it is homoplastic (Leaw et al. 2005).
Using the classical species definition for lineage formation (Mayr 1982), Lilly et al. (2007) recognize Alexandrium tamarense (Lebour) Balech, Alexandrium catenella (Whedon and Kofoi) Balech and Alexandrium fundyense Balech as different species. On the other hand, phylogenies of Alexandrium species have been established based on genomic sequences of the large and small subunits of ribosomal DNA (LSU and SSU rDNA, respectively) (Guillou et al. 2002;Usup et al. 2002;John et al. 2003John et al. , 2005Murray et al. 2005;Rogers et al. 2006). Of these sequences, the D1/D2 region of the LSU rDNA has been proved to be the most suited for discrimination of closely related Alexandrium species (Ki and Han 2007). Thus, this hypervariable region has been proposed as a suitable candidate to discriminate between species with similar fidelity as Cytochrome Oxidase I gene (Sonnenberg et al. 2007). Anderson (1994, 1996) and Scholin et al. ( , 1995, based on DNA sequencing of the divergent D1/D2 LSU rDNA region and restriction fragment length polymorphism (RFLP) analysis of the small subunit rDNA genes, consider that they could be strains of the same species, naming them the tamarensis complex. Using these results as a starting point, Lilly et al. (2007) further established the tamarensis complex as a valid cluster, derived from a phylogenetic analysis comparing more than 126 different Alexandrium D1/D2 region sequences. Their detailed examination revealed the presence of five clades, defined as: Group I (North American), Group II (Mediterranean), Group III (Western European), Group IV (Temperate Asian) and Group V (Tasmanian).
In this study, we sequenced local Alexandrium intergenic spacer 1 (ITS1), 5.8S rDNA, ITS2 and D1-D5 hypervariable LSU rDNA regions in order to incorporate molecular data that could help define more clearly the Alexandrium species responsible for HABs in Chilean coasts. This is considering that the local South American Pacific Alexandrium species were classified as A. catenella (Muñ oz 1985), mainly based on morphological traits, but without a further assessment of sequence variation. In this sense, speciesspecific assignment allows the implementation of a polymerase chain reaction (PCR) assay for accurate monitoring along Chilean coasts in order to prevent public health hazards and economic losses.

DNA purification from cell cultures
Cells for analysis (100 mL) were collected from each clonal culture at mid-logarithmic phase and centrifuged at 3000 g for 5 min at 4 8C. The supernatant was removed, the pellet resuspended in 500 mL of Milli-Q water, and transferred to a 1.5-mL microfuge tube. Microfuge tubes were placed in liquid nitrogen for 30 s and the cells subsequently disrupted for 1 min using an Axygen polypropylene pestle (PES-15-B-SI, Union City, CA, USA). Genomic DNA was then extracted from the pellet using the DNeasy Plant Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions.
Determination of Alexandrium DNA concentration and quality DNA concentration was measured using a fluorometer (Qubit, Sunnyvale, CA, USA) together with the Qubit dsDNA BR Assay Kit (Invitrogen, Eugene, OR, USA). The quality was evaluated based on its integrity by comparison with a 23-kb band of l-HindIII ladder (Invitrogen, USA) in 1 % agarose (Ultrapure, Invitrogen, Barcelona, Spain) gel electrophoresis stained with ethidium bromide.

PCR amplification and sequence analysis
In order to determine the species of the three isolates, the ITS1-D5 rDNA was amplified, sequenced and then aligned with sequences from GenBank (Ki and Han 2007). All amplifications were carried out in duplicate with 1× PCR buffer, 20 -50 ng of genomic DNA template, 3 mM MgCl 2, 100 mM each dNTP, 0.1 mM each primer and 0.4 U of recombinant Taq DNA polymerase (MBI Fermentas, Vilnius, Lithuania) in a 10-mL reaction volume. Polymerase chain reaction primer sequences for LSU rDNA and optimized annealing temperatures are specified in Table 1. Polymerase chain reaction parameters were: 95 8C for 5 min; 35 cycles of denaturation at 95 8C for 30 s, annealing for 30 s, extension at 72 8C for variable time spans which depended upon the size of the amplifying fragment (1000 bases min 21 ); and a final extension at 72 8C for 5 min. Reactions were run on a MaxyGene Gradient thermocycler (Axygen). Five microlitres of PCR products were analysed by 2 % agarose (Invitrogen, USA) gel electrophoresis according to standard methods. rDNA PCR amplification products from clones ACC01, ACC02 and ACC07 were purified from gels using the MinElute Gel Extraction Kit (Qiagen) following the manufacturer's instructions  The annealing temperature for each primer and the study from which the sequences were obtained are listed. and directly sequenced in an ABI PRISM 3100. An electropherogram base quality assignment algorithm, phred , was used to reanalyse the sequenced fragment of all strains in order to determine intragenomic polymorphic sites. Additional confirmation of species identification was achieved by amplifying the extracted DNA using four microsatellite primers specific to A. catenella (Nagai et al. 2005; Table 2). The amplification conditions were the same as those provided in the original publications describing the assays.

Collection, analysis and association of sequences
All selected Alexandrium sequences were obtained using the keywords 'Alexandrium LSU rDNA' or 'Alexandrium 28S in the GenBank database (http://www.ncbi.nlm.nih. gov) [see ADDITIONAL INFORMATION]. For a detailed species-specific analysis, two data sets were generated. The first was composed of 81 unique sequences at least 641 bp long, covering the D1/D2 region of the LSU rDNA. This group incorporated 79 Alexandrium genus species (including the local strain), and two Prorocentrum micans strains that were used as outgroups. The second set had 18 sequences at least 1776 bp long, of which 16 corresponded to tamarensis complex (including the local strain), one to A. minutum and one to A. affine; the last two were used as outgroups. For both data sets, alignments were carried out in the ClustalX2 V2.0 (Larkin et al. 2007) graphical platform.

Substitution model and associated parameter estimation
jModeltest (Posada 2008) was used to find the best substitution model and associated parameters for phylogenetic analysis in both data sets using the Akaike (Hirotugu 1974) and Bayesian (Schwarz 1978) information criteria.

Phylogenetic analysis
Bayesian analysis was implemented with MrBayes V3.2 (Ronquist et al. 2012) for the first and second data sets, and was carried out with 1 500 000 runs, with five separate initial trees, with the Markov chain Monte Carlo (MCMC) process set to four chains and 25 % of initial trees discarded as 'burn-in'. Within each chain, samples were obtained every 100 iterations, and the values of the average deviation of split frequencies (AVSF) and potential scale reduction factor (PSRF) were obtained. These values were used to evaluate convergence of the generated trees. Additionally, maximum

ATF11-R acctgcggctgcgacacgact
The annealing temperature for each primer, the species and the study from which the sequences were obtained are listed. TD, touchdown.
likelihood analysis was carried out in PhyML V3.0 (Guindon and Gascuel 2003) in order to further support taxon assignment. Analysis was started with a random tree sample (Subtree pruning and regrafting method) and 1000  In order to estimate the nucleotide differences within the tamarensis complex in each data set, the values for the average number of differences per pair of sequences aligned within each group and the number of parsimonious informative sites were calculated in the MEGA 5.05 program (Tamura et al. 2011). Alexandrium DNA detection in challenged Mytilus samples by real-time PCR An in situ experimental protocol for dinoflagellate challenge was developed in order to implement it subsequently for Alexandrium detection in filter-feeding shellfish and water columns. The experiments were performed in quadruplicate using four experimental aquaria of 15 L, containing five mussels each (one mussel from each aquarium for each of the 5 days was used for DNA extraction).
Individuals of the edible mussel Mytilus were transported to the laboratory, where they were acclimatized for 1 week at 14 8C and seawater salinity of 30 practical salinity units. During this period, mussels were continuously fed with the microalga Isochrysis galbana at 1.5 mg L 21 using a peristaltic pump and providing constant aeration. The seawater was changed every 48 h. Following acclimation, the mussels were exposed to a contaminated diet (1.7 -2.0 mg L 21 ; dry weight) containing 50 % toxic dinoflagellate Alexandrium strain ACC02 and 50 % I. galbana (by weight) for a period of 12 days, followed by a detoxification period of 15 days where they were fed with I. galbana. Every day the aquaria received an amount of food representing 2 % Fig. 1 Amplification of local Alexandrium strains ACC01, ACC02 and ACC07 using species-specific primers; A. tamarense (lanes 1, 2 and 3) and A. catenella (lanes 4, 5 and 6). M ¼ 100-bp DNA size marker. Species-specific amplification in the rDNA region using A. catenella and A. tamarense primers were carried out in a MaxiGene Gradiente thermocycler (Axygen) in 1× PCR buffer, 20-50 ng of genomic DNA template, 3 mM MgCl 2, 100 mM each dNTP, 0.1 mM each primer and 0.4 U of TopTaq DNA polymerase (Fermentas) in a 10-mL reaction volume. Five microlitres of each PCR product were analysed in a 2 % agarose gel. A Fermentas GeneRuller TM 100-bp DNA ladder was used for size estimation of amplified fragments. Fig. 2 Alexandrium tamarense microsatellite amplifications of the three local strains with (A) specific primers ATB8 (lanes 1, 2 and 3) and ATD8 (lanes 4, 5 and 6). Lanes 7 and 8 correspond to controls with primers ATB8 without DNA. (B) Specific amplification with primers ATB1 (lanes 1, 2 and 3) and ATF11 (lanes 6, 7 and 8). Lanes 4-5 and 9 -10 are controls without DNA for primer sets ATB1 and ATF11, respectively. M ¼ 100-bp DNA size marker. Species-specific microsatellite amplifications using A. catenella and A. tamarense primers were carried out in a MaxiGene Gradiente thermocycler (Axygen) in 1× PCR buffer, 20 -50 ng of genomic DNA template, 3 mM MgCl 2, 100 mM each dNTP, 0.1 mM each primer and 0.4 U of TopTaq DNA polymerase (Fermentas) in a 10-mL reaction volume. Five microlitres of each PCR product were analysed in a 2 % agarose gel. A Fermentas GeneRuller TM 100-bp DNA ladder was used for size estimation of amplified fragments. of the dry body weight of the experimental mussels (Shafee 1976), delivered continuously using a Masterflex 7519-05 peristaltic pump at the temperature and salinity cited above. One mussel from each replicate aquarium was taken on Days 2, 3, 4 and 5 of the toxic feeding cycle and on Day 15 of the detoxification cycle. Animals were immediately processed for purification of gill DNA.
To determine the possibility of detecting Alexandrium DNA in challenged mussel tissue, we used real-time PCR. Experiments were run in triplicate, for each sample, on an Eco real-time PCR System (Illumina, San Diego, CA, USA) using Quantace SensiMix HRM tm kit (Bioline, London, UK). Reaction conditions were: 1 × Sen-siMix HRM buffer, 0.6 mL of EvaGreen dye, 0.5 mM primers tamF and tamR (Hosoi-Tanabe and Sako 2005), and 100 ng of challenged Mytilus spp. gill DNA in a 10-mL final reaction volume. The PCR protocol cycling was: a 10-min initial activation step at 94 8C, followed by 40 cycles of 94 8C for 30 s, 55.3 8C for 30 s and 72 8C for 30 s. A PCR amplification product of 235 bp (Fig. 1) obtained with tamF and tamR was purified and sequenced in order to corroborate specificity.

DNA purification from challenged Mytilus gill tissue
Mussels were randomly taken from each of the four aquaria (replicates) on each day of sampling. Animals were dissected alive and 30 g of drained gill tissue were used as the starting material for DNA purification with DNeasy Blood & Tissue Kit (Qiagen) according to the manufacturer's protocol. Each sample of purified DNA was stored at 220 8C. As explained below, the gill was considered as a useful source of Alexandrium DNA as particulate materials, such as unicellular organisms, tend to accumulate in this organ (Jørgensen 1996;Riisgard et al. 1996) and other tissues such as the hepatopancreas do not provide DNA with the integrity needed for this type of study.

PCR amplification of microsatellite genomic regions
The results of the species-specific PCR of eight microsatellite genomic regions were consistent in all three isolates, amplifying only those directed towards A. tamarense and not to A. catenella (Fig. 2). No changes could be observed with the A. catenella set of primers despite the numerous protocol modifications of PCR conditions. Analysis based on ITS1-D5 LSU rDNA sequences Sequence analysis and evaluation Electropherogram profiles from the ITS1-D5 region of the rDNA of strains ACC01, ACC02 and ACC03 were analysed with the phred algorithm in order to discard the presence of pseudogenes or intragenomic rDNA polymorphisms (IRP). Only bases with scores over 30 (sequencing error probability 1/1000) were considered for further analysis. As no nucleotide differences were obtained for this region between the three local Alexandrium strains, only one sequence, Ach01, was used as a representative of them (NCBI accession no. JN657223).
Substitution model and associated parameter evaluation Analysis by jModeltest estimated that the GTR + G was the best substitution model for later Bayesian and maximum likelihood analysis. If a given model was not an option in MrBayes, the following least restrictive model was used (e.g. GTR).
Phylogenetic analysis Using 81 sequences of length 641 bp, all aligning in the same D1/D2 LSU rDNA region, from different species of the genus Alexandrium and two strains of P. micans, the phylogenetic distribution of local Alexandrium strains could be estimated (Fig. 3). Convergence of Bayesian trees was evaluated through AVSF and PSRF. Values were less than 0.01 and 1 +0.005, respectively, suggesting that the distribution reached a stationary phase in both data sets. Additionally, bootstrap values and the logarithm of the likelihood score of the optimal tree (24229.43342) extracted by maximum likelihood analysis were indicative of a precise tree. The results indicate that the local Alexandrium strain is in Group I, dominated by A. tamarense, and is grouped with other previously sequenced Alexandrium strains from Chilean waters (ACC01, ACC02 and ACC07; Fig. 3).
In order to achieve further resolution of the tamarensis complex, 18 sequences longer than 1776 bp, from Fig. 3 Phylogenetic tree of 80 Alexandrium species and strains (Atm, A. tamarense; Act, A. catenella; Afu, A. fundyense; Amn, A. minutum; Aaf, A. affine; Atr, A. tropicale; Aty, A. tamiyanavichii; P. micans, Prorocentrum micans) and local Alexandrium species Ach01. Sequences were obtained from the GenBank database using the keywords 'Alexandrium LSU rDNA' or 'Alexandrium 28S'. Phylogenetic trees were generated for an alignment of 81 sequences of 641 bp in the D1/D2 region through Bayesian inference in MrBayes V3.2 and maximum likelihood (ML) in PhyML. Bayesian analysis was carried out with 1 500 000 runs, with five separate initial trees. Convergence was checked through PSRF and AVSF. FigTree V1.3.1 was used as a visual representation of output trees. For ML analysis these were carried out with 1000 bootstrap replicates. different species and strains of the tamarensis complex, located in the rDNA region were analysed (Fig. 4). The total alignment involved ITS1, 5.8S rDNA, ITS2 and 1185-bp of the D1 -D5 regions of the 28S rDNA (Ki and Han 2007). Sequences of A. minutum and A. affine were selected as outgroup species for the tamarensis complex.
For maximum likelihood analysis, the logarithm of the likelihood score of the optimal tree was 26591.25984. As expected, the results were again consistent and the local Alexandrium strain was allocated to Group I of the tamarensis complex.
Topologies for trees generated through maximum likelihood and Bayesian inference were the same, and had high bootstrap and posterior probability values (Figs 3 and 4). Group generation was carried out analysing clades formation and the previous literature. Comparing the structure of both trees, using 18 or 81 sequence alignments, the same clades were formed.
In order to obtain information on the variability of the amplified region for both data sets, the average number of differences per pair of sequences aligned and the number of parsimonious sites were measured. For the first data set, values were 8.37 bp and 130, respectively. On the other hand, for the second data set, the average number of differences per pair of sequences aligned was 52.03 bp and the number of parsimonious sites was 298. Considering the 1185-bp segment covering only the LSU rDNA region, the amount of parsimonious informative sites and the average number of differences per pair of Fig. 4 Phylogenetic tree of 17 Alexandrium species and strains (Atm, A. tamarense; Act, A. catenella; Afu, A. fundyense; Amn, A. minutum; Aaf, A. affine) and local Alexandrium species Ach01. Sequences were obtained from the GenBank database using the keywords 'Alexandrium LSU rDNA' or 'Alexandrium 28S'. Phylogenetic trees were generated for an alignment of 18 sequences of 1776 bp in the ITS1-D5 region through Bayesian inference in MrBayes V3.2 and maximum likelihood (ML) in PhyML. Bayesian analysis was carried out with 1 500 000 runs, with five separate initial trees. Convergence was checked through PSRF and AVSF. FigTree V1.3.1 was used as a visual representation of output trees. For ML analysis these were carried out with 1000 bootstrap replicates. sequences aligned decreases from 298 to 174 and from 52.02 to 35 bp, representing a fall of 42 and 32 %, respectively, with respect to the whole amplified region.
Alexandrium DNA detection in challenged Mytilus samples by real-time PCR From each aquarium, DNA from the gill tissues of five different randomly picked Mytilus challenged in vivo with Alexandrium strain ACC02 were used to detect the presence of A. tamarense through a real-time PCR assay with species-specific primers. We obtained similar and positive results for the samples, which were extracted between Days 2 and 5 of the toxicfeeding phase, with Ct values ranging from 21 to 22. On the other hand, all samples extracted on Day 15 of the detoxification phase gave no specific amplification, indicating that there was no detectable Alexadrium DNA in the Mytilus samples (Table 3). Replicates of Mytilus samples challenged in parallel in four independent aquaria showed the same pattern, confirming the detection of Alexandrium in Mytilus gills during the toxic phase. Through amplicon melt analysis and direct sequencing of the 235-bp PCR-amplified fragment, the region corresponded to the expected specific sequence within the D1/D2 LSU rDNA domain, according to Hosoi-Tanabe and Sako (2005).

Discussion
The study of HABs has become increasingly important given the recent rise in the number and frequency of toxic events, and associated adverse impacts on public health, fisheries and ecosystem services (Anderson 1989;Cassis et al. 2002). This threat has led to extensive investigation on how to develop international standards for the detection of toxins in seafood and on the implementation of expanded monitoring programmes for toxic algae. One of the most frequently used methods for the evaluation of toxins in HAB episodes worldwide has been the mouse bioassay. Unfortunately, this technique does not always provide accurate estimates of toxicity (Ferná ndez et al. 2002) and requires considerable resources and time. Moreover, this technique has been questioned with regard to animal welfare and is prohibited in some countries. In this context, increasing efforts have been focused on monitoring the toxic algae directly to predict their occurrence and better allocate sampling effort, particularly with regard to toxin analysis.
In this paper, we present phylogenetic analyses using ITS1-D5 rDNA sequence data which demonstrate that the Chilean strains analysed belong to Group I in the tamarensis complex, rather than to the A. catenella grouping Medlin et al. 1998;Higman et al. 2001;Lilly et al. 2007). The first studies concerning Chilean HABs carried out 40 years ago, based on morphological observations, identified A. catenella as the dominant Alexandrium species (Muñ oz 1985). This identification has never been questioned or assessed by more accurate methods such as genomic sequencing. The phylogenetic analysis carried out in this study clearly indicates that the local Alexandrium species belongs to the Group I ribotype. This group is mainly composed of A. tamarense, in contrast to Group IV in which the predominant species is A. catenella. Similarly, phylogenetic trees based on Alexandrium toxin variability showed that strains from Argentina, Brazil, Chile and Uruguay belonged to the same clade, paralleling the Group I results (Montoya et al. 2010). As red tide blooms have been present since the 19th century in Chile and Brazilian, Uruguayan and Argentinian events are more recent, it has been proposed that toxic episodes in Eastern South America could be due to the expansion of Chilean Alexandrium species (Lilly et al. 2007). Even more, Uruguayan and Brazilian strains have been classified as A. tamarense in Group I, consistent with our findings in relation to the fact that Chilean species, supporting the hypothesis that the local strain was missclassified as A. catenella.  rDNA subunits as a suitable method for strain differentiation in Alexandrium species, due to the existence of paralogue genes. In this respect, base quality discrimination did not give evidence of paralogue sequences. On the other hand, Ki and Han (2007), eliminating paralogue sequences for their analysis, found 39 parsimony informative sites within the D1 -D5 LSU rDNA region in five different Alexandrium species. Our study found a much higher value of parsimony variable sites (174) when our local sequence was aligned with published sequences, which probably contained paralogue DNA regions. It is unlikely that this considerable difference could be explained by increased mutation rates in the sequenced regions. Thus, this result agrees with Miranda et al. (2012), who suggest that the great diversity in Group I of the tamarensis complex could be due to the lack of an accurate discrimination of paralogue sequences.
The sequence analysis facilitated the use of a specific and highly sensitive real-time PCR assay to detect local Alexandrium. Owing to the ability of filter-feeding molluscs to capture and concentrate phytoplankton, by pumping water through their gill filaments, we tested the possibility of detecting dinoflagelate DNA in this organ of challenged mussels. Preliminary experimental results showed, for the first time, the implementation of a practical test to detect these algae in gill tissue extracted from mussels challenged under laboratory conditions. Currently, we are working on the implementation of this test in field samples in order to detect traces of Alexandrium in seawater. It would be very useful to count with methods to detect traces of dinoflagellates, in order to predict massive algal blooming through constant monitoring of red tide episodes, thus preventing human consumption of toxic filter-feeding shellfish.

Conclusions and forward look
Traditionally, South American Pacific HABs have been assigned to A. catenella, based only on morphological evidence that has proven to be an unreliable indicator of species identification within the A. tamarense complex (Lilly et al. 2007).
This study was focused on the molecular identification of the Alexandrium species that causes paralytic shellfish poisoning in Chilean coasts (Herná ndez et al. 2005). Phylogenetic analyses based on sequence data and species-specific PCR assays targeting LSU rDNA and microsatellite regions, all demonstrate that cultures isolated from Chilean coasts belong to the tamarensis complex Group I and are not A. catenella (Figs 1 and 2, Tables 1 and 2, Hosoi-Tanabe and Sako 2005).
As not all A. tamarense are toxic, we are currently developing a real-time PCR assay based on primer pairs that target signature nucleic acid sequences of genes involved in toxin production. Our goal is to set up a new technique for early, sensitive and accessible HAB detection in order to avoid the important financial damage and public health issues.

Additional information
The following additional information is available in the online version of this article -Sequences used for phylogenetic analyses. File 1: DNA sequences used in this study. Strain assignment, morphospecies, origin and accession number in GenBank are given when available.

Accession numbers
The Ach01 (NCBI accession no. JN657223) sequence was uploaded to GenBank.

Sources of funding
This work was funded by the Corporació n del fomento de la producció n (Corfo) of Chile through the INNOVA Chile Project #07CN13PPD-240.