• 1

    Received 23 April 2008. Accepted 25 July 2008.


Karlodinium veneficum (D. Ballant.) J. Larsen strains, 16 from the U.S. Atlantic eastern seaboard and two from New Zealand (CAWD66 and CAWD83), were used to characterize toxin profiles during batch culture. All 18 strains were determined as the same species based on ITS sequence analyses, a positive signal in a chloroplast real-time PCR assay and pigment composition. Five karlotoxin 1 (KmTx 1) containing strains were analyzed from the Chesapeake Bay, and 10 karlotoxin 2 (KmTx 2) strains were analyzed from Florida to North Carolina. One strain (MD5) from the Chesapeake Bay produced no detectable toxin. The two cultures from New Zealand contained both novel karlotoxins with lower masses and earlier elution times. Toxin type did not change during batch culture, although the KmTx phenotype did change in some strains under extensive (months) phototrophic growth in replete media. KmTx cell quota did not change during batch culture for most strains. The mass spectrum for every KmTx examined showed a pattern of multiple coeluting congeners within each HPLC peak, with masses typically differing by 16 amu. KmTx congeners tested showed nearly a 500-fold range in specific hemolytic activity, with KmTx 1 (typically occurring at lower cellular levels) most hemolytic and CAWD66 toxin least hemolytic, while KmTx 2 and the CAWD83 toxin had similar intermediate specific activity. Despite morphological, genetic, and photopigment indicators consistent with species homogeneity among the 18 strains of K. veneficum, the high degree of toxin variability suggests different functional roles among strains that likely coexist in situ.


Cawtthron Culture collection


Culture Collection of Marine Phytoplankton


cetyl trimethylammonium bromide


deoxyribonucelotide triphosphate


enriched seawater artificial water


equivalent spherical diameter


equivalent spherical volume




Liquid Chromatography–Mass Spectrometry




tetrabutyl ammonium acetate


Tris-EDTA buffer

K. veneficum is a cosmopolitan temperate estuarine dinoflagellate responsible for numerous fish kills on the eastern seaboard of the United States (Deeds et al. 2002, Kempton et al. 2002, Fensin 2004, Goshorn et al. 2004, Hall et al. 2008). Previously, K. veneficum was called Gyrodinium estuariale, Gymnodinium galatheanum, Gymnodinium veneficum, and K. micrum, all of which are now synonymous with K. veneficum (Bergholtz et al. 2005). Much of K. veneficum’s ecology, such as its ability to capture prey (Adolf et al. 2008b), avoid grazing (Adolf et al. 2007), affect co-occurring algae (Adolf et al. 2006), or be parasitized (Bai et al. 2007), is mediated by the production of a unique polyketide toxin, putatively called karlotoxin (KmTx) (Deeds et al. 2002, Bachvaroff et al. 2008a). There is a geographic cline in karlotoxin production (Deeds et al. 2004) along the U.S. eastern seaboard. K. veneficum strains from points south of the Chesapeake Bay produce karlotoxin 2 (KmTx 2), whereas strains from within the Chesapeake Bay produce karlotoxin 1 (KmTx 1), suggesting discrete genetic populations, although the genetic identity for these populations has only recently been investigated (Bachvaroff et al. 2008b).

Strain or species variation in toxin production has been described for many harmful algae including the dinoflagellate genera Gambierdiscus (Bomber et al. 1989), Alexandrium (Cho and Lee 2001), and Gymnodinium (Blackburn et al. 2001); the diatom Pseudo-nitzschia (Scholin et al. 1994, Evans et al. 2004); the cyanobacteria Microcystis (Wilson et al. 2005) and Nodularia (Bolch et al. 1999). Similarly, genetic variability in growth rates for dinoflagellates has been documented, although one species, Prorocentrum micans, had identical growth rates among its clones (Costas 1990). For a globally distributed ichthyotoxic phytoplankton such as K. veneficum, understanding the relationship between genetic (ribotype) and phenotypic (i.e., toxicity) variability is important to understanding the environmental distribution of toxic blooms and fostering the development of improved tools for monitoring and predicting these events.

Here, the growth rate, pigments, and ITS ribotypes were determined for 18 K. veneficum clonal isolates. In addition, the amount and type of karlotoxin in these same strains was monitored during a typical culturing cycle of 28 d from inoculation to declining phase. Most of these strains (16) are from the U.S. eastern seaboard, including six from the Chesapeake Bay and 10 from south of the Bay, with two New Zealand isolates as a geographic outgroup. Of these strains, 11 had not yet been characterized for toxin type or amount (Deeds et al. 2004), and the toxins produced by the New Zealand strains have not been characterized. The current survey demonstrates significant variation in toxin type, toxin cell quota, and pigments for both regional U.S. populations and the two New Zealand isolates grown under identical conditions. Surprisingly, very few growth rate differences were observed. The results provide insight into the extensive variation in phenotypic characters for K. veneficum strains, even when grown under identical conditions and question whether we are dealing with cryptic species within natural K. veneficum populations.

Materials and methods

Culturing. K. veneficum strains were acquired from the Provasali-Guillard Center for the Culture of Marine Phytoplankton (CCMP, Andersen et al. 1997), the Cawthron Culture collection (CAWD), M. Johnson of Horn Point Labs, and S. Kibler of NOAA (Table 1). The algae were grown in 15 ppt natural seawater (Indian River inlet, DE, USA) combined with f/2 nutrients and vitamins without Si (Andersen et al. 1997) at 100 μm photons · m−2 · s−1 PAR (14:10 light:dark [L:D]), 20°C, unless otherwise noted. Growth irradiance was measured below cool-white fluorescent lamps with a Li-Cor QUANTUM probe attached to a Li-Cor LI-250 light meter (Li-Core Biosciences, Lincoln, NE, USA). These clonal cultures were not axenic, and the growth conditions were not necessarily optimum for each strain. For each strain examined, three replicate batch cultures were grown. Cell abundance and equivalent spherical diameter (ESD) were measured on a Coulter® Counter (Beckman Coulter, Fullerton, CA, USA) using the “narrow” size range (4–30 μm). Equivalent spherical volume (ESV) was calculated as (4/3)πr3, where r is the radius derived from ESD. Either a 500 or 2,000 μL sample was counted. Samples that yielded a “coincidence correction” value >20% were diluted 1:10 with sterile seawater at salinity 15.

Table 1. Karlodinium veneficum strains examined.
Culture nameToxin typeClonalCulture sourceCollection siteDateCell diameter (μm) Division rate (μ)Maximum density (cells · mL−1) × 103Toxin cell quota (pg · cell−1)PCR CTFish kill
  1. Superscripts denote higher (a) to lower (d) average values for a group.

  2. *Bimodal distribution with means at 8.9 and 11.5 μm.

  3. †Measured in ISOTON, which swells the volume 16 ± 2.3% relative to growth media.

  4. ‡Based on 20 ng genomic DNA input.

MD2KmTx 1YesHPL38.613 N 76.146 W Choptank River, MDMarch 20, 19999.21 ± 1.3940.23 ± 0.007a200.9 ± 21.45a,b0.06 ± 0.016d22.3 ± 0.12 
MD5NoneYesHPL38.613 N 76.146 W Choptank River, MDMarch 20, 19998.29 ± 1.1960.25 ± 0.007a260.8 ± 23.8a<0.00120.3 ± 0.05 
MD6KmTx 1YesHPL38.613 N 76.146 W Choptank River, MDMarch 20, 19999.52 ± 1.320.21 ± 0.008a122.5 ± 14.54b0.04 ± 0.017d20.5 ± 0.36 
CCMP 1974KmTx 1YesCCMP37.85 N 76.1 W Chesapeake Bay, VAMay 199510.83 ± 1.4870.25 ± 0.027a148.1 ± 8b0.29 ± 0.037d18.9 ± 0.16 
CCMP 1975KmTx 1YesCCMP38.1733 N 75.7374 W Hyrock Farms, MDJuly 199610.20 ± 1.360.26 ± 0.139a138.7 ± 8.2b0.05 ± 0.002d20.3 ± 0.23+
SlocumKmTx 2YesNOAA, Steve Kibler35.951 N 76.9 W Slocum Creek, NC October 6, 200310.02 ± 1.3540.21 ± 0.038a243.7 ± 16.52a1.06 ± 0.233b21.1 ± 0.06 
FB3KmTx 2YesNOAA, Steve Kibler34.984 N 76.95 W W Flanner’s Beach, Neuse River, NCOctober 7, 20039.61 ± 1.1940.18 ± 0.078a53.2 ± 15.48c1.89 ± 0.795b19.9 ± 0.21 
MBM1KmTx 2YesNOAA, Steve Kibler34.968 N 76.818 W Minnesott River Marina, Neuse River, NCOctober 8, 200310.08 ± 1.5380.27 ± 0.010a121.7 ± 5.28b0.60 ± 0.047b21.5 ± 0.30 
CCMP 2388KmTx 2YesCCMP32.83 N 79.863 W Hobcaw Creek, SCFebruary 5, 200210.75 ± 1.6720.30 ± 0.101a108.8 ± 22.48b0.04 ± 0.015d20.1 ± 0.24+
IB4KmTx 2YesSERC32.2167 N 80.7335 W Hilton Head, SCFebruary 5, 200210.66 ± 1.2230.35 ± 0.049a203.7 ± 40.11a0.42 ± 0.169c23.3 ± 0.08+
CCMP 2282KmTx 2YesCCMP32.2167 N 80.7335 W Hilton Head, SCMarch 30, 200110.72 ± 1.2640.29 ± 0.09a235.4 ± 11.39a1.39 ± 0.504b19.1 ± 0.02+
CCMP 2283KmTx 2YesCCMP32.2167 N 80.7335 W Hilton Head, SCMarch 30, 200110.69 ± 1.3810.36 ± 0.06a196.0 ± 10.08b0.63 ± 0.109b21.2 ± 0.19+
CCMP 2064KmTx 2YesCCMP31.9537 N 81.0058 W Wilmington River, GANovember 9, 199810.14 ± 1.2320.22 ± 0.013a193.9 ± 4.50c2.0 ± 0.51b19.9 ± 0.09+
PD-6KmTx 2YesG. Smalley31.96 N 80.938 W Wassaw Island, GAJune 2, 20029.08 ± 1.2670.32 ± 0.009a238.5 ± 18.52b0.56 ± 0.072b19.6 ± 0.16 
F4KmTx 2YesNOAA, Steve Kibler30.4145 N 81.5285 W St. John’s River, FLApril 4, 200110.81 ± 1.86*0.33 ± 0.036a58.1 ± 1.71d3.14 ± 0.619a25.0 ± 0.13 
CCMP 2778KmTx 2YesCCMP27.3122 N 82.6010 W Sarasota, FLFebruary 28, 200511.90 ± 1.330.24 ± 0.029a34.8 ± 5.94d4.28 ± 1.74a17.7 ± 0.10 
CAW D66KmTx 2-likeYesCawthronUnknown199312.07 ± 1.678†0.23 ± 0.040a110.9 ± 9.47c0.38 ± 0.008c25.0 ± 0.06 
CAW D83KmTx 1-likeYesCawthron41.283 S 174.167 E Whangakoko199410.78 ± 1.534†0.17 ± 0.065a200.8 ± 21.24b,c0.20 ± 0.010d23.5 ± 0.13 

Instantaneous growth rate for each sampling interval of individual growth curves was determined as follows:


where N0 and N1 are the cell densities at the beginning and end of each sampling interval, respectively, and t is time in days. Growth rate is expressed here as d−1 and can be converted to divisions per day by dividing the growth rate by Ln 2 = 0.69. Maximum cell yield was determined for individual growth curves by inspection of cell densities determined on individual sampling days. The strain average for maximum growth rate and maximum cell yield was calculated using values from each replicate growth curve (n = 3 per strain).

Light microscopy.  Each strain was viewed and photographed at ×1,000 using a Zeiss Axiscope equipped with a Zeiss Axiocam interfaced to a computer running AxioVision software (Carl Zeiss Optical Inc., Chester, VA, USA).

Molecular genetic characterization.  Cell pellets from 105 to 106 cells were obtained by centrifugation (IEC Centra CL2, Thermo Fisher Scientific, Waltham, MA, USA) at 3,000g (∼1,600g) and frozen at −20°C prior to DNA extraction. DNA was isolated using a CTAB detergent method followed by chloroform extraction and precipitation of the aqueous phase with isopropanol (Doyle and Doyle 1987). The DNA was diluted based on 260 nm absorbance, and 100 ng was used in a 20 μL PCR reaction in the following buffer: 3 mM MgCl, 500 mM Tris-HCl pH 8.3, 500 μg · mL−1 BSA, 2 mM dNTPs, and 4 pmoles of primer for 35 cycles (all concentrations final). The temperature was cycled from a 15 s denaturating at 94°C to a 55°C annealing for 15 s followed by 30 s of extension at 72°C for 35 cycles. The primers used were based on primers suggested by Litaker (Litaker et al. 2003): ITSFor CTGCGAAGCTATCGCTATT and ITSRev TGAGGGAATCCTATTTAG designed to cross from ITS1 to the LSU rRNA including most of ITS1, 5.8S, and ITS2. PCR amplicons were purified with equal volumes of 20% w/v polyethylene glycol (mw 8000) containing 2.5 M NaCl, pelleted by centrifugation, and washed with 70% ethanol. Sequencing reactions were performed and run on an ABI 3130XL genetic analyzer according to the manufacturer’s protocol. Sequences were edited using Sequencher 4.5 software (Genecodes Corp., Ann Arbor, MI, USA) and exported to MacClade 4.05 (Sinauer Associates Inc., Sunderland, MA, USA) for manual alignment using GenBank sequences AF352365–7 for K. veneficum as a reference.

Given the greater variability in the plastid 16S SSU gene sequence (Tengs et al. 2001), the TaqMan detection assay was also used to test the identity of these strains using the forward primer GCAACCCTTGTTTGGTCAG, reverse primer AGTAAGCGGCTCTTTGTCTTAACC, and probe TGAGAAATCGGAGGAAGGTAAGGATGACG (Tengs et al. 2001). All DNA extractions were completed by spinning down 25–50 mL of a culture, which was resuspended in CTAB solution (2% hexadecyltrimethlyammonium bromide, 0.7 M NaCl, 10 mM EDTA, 50 mM Tris pH 8.0) and frozen at −20°C. The pellets were thawed and mixed with 1 mL hot CTAB, then pipetted into a sterile 2 mL tube and heated in a hot water bath at 65°C × 10 min. Samples were allowed to cool to room temperature before adding 1 mL chloroform:isoamyl alcohol solution to each. Samples were centrifuged at 15,000g × 10 min, the aqueous phase was removed, then an equal volume of isopropanol was added to the sample, which was centrifuged at max. g × 15 min. The isopropanol was poured off, and 1–2 mL 70% ethanol was added to the sample. The sample was centrifuged at maximum speed for 5 min, the ethanol was poured off, and the pellet was allowed to dry. Dried pellets were resuspended in 100 μL TE (or H2O).

The nucleic acid concentration of each sample was determined using the NanoDrop-1000 spectrophotometer (Thermo Fisher Scientific) according to the manufacturer’s recommendations. Purified DNA samples were diluted to concentrations of 10 ng · μL−1 working dilutions using the formula C1 × V1 = C2 × V2. Working dilutions were run according to the manufacturer’s directions using TaqMan® Universal PCR Master Mix (Applied Biosystems, Foster City, CA, USA) in MicroAmp Optical 96-well reaction plates (Applied Biosystems) on an ABI PRISM® TaqMan® 7700 Sequence Detection System (Applied Biosystems) under the standard real-time PCR conditions except for the sample volume, which was changed to 20 μL (50°C × 2 min, 95°C × 2 min, 95°C × 2 min, [60°C × 1 min] × 40 cycles). Sterile dH2O was used as the background component of every sample. DNA from K. veneficum strain CCMP 2064 was used to create the standard curve dilution series. The standard curve dilution series was run in all TaqMan® assays of unknown purified DNA. To create the standard curve dilution series, the stock was serially diluted to final DNA concentrations of 20 ng · μL−1, 2 ng · μL−1, 200 pg · μL−1, 20 pg · μL−1, and 2 pg · μL−1. The total DNA concentration of the standard was brought up to 20 ng (in a total sample of 10 μL) by the addition of P. piscicida DNA, which has been previously shown to be a good negative control for the primers and probe used in this study (Tengs et al. 2001). P. piscicida DNA (20 ng · μL−1) was used as both the negative control and 0 ng · μL−1 K. veneficum DNA in the standard curve dilution series. The standard curve dilution series was assayed in triplicate. The mean CT values of the assay were plotted against the log DNA concentration so that a linear regression could be calculated from the data points. DNA from each strain (20 ng) was run in triplicate and compared to the standard curve. Typical standard curve efficiencies exceeded 90% with a lower detection level of 2 pg of template.

Pigment analyses.  Pigments were analyzed according to Van Heukelem and Thomas (2001) using a C8 HPLC column (Agilent Eclipse XDB-C8, part number 963967-906, Agilent Technologies, Santa Clara, CA, USA) coupled to a methanol-based reversed-phase gradient solvent system using tetrabutyl ammonium acetate (TBAA, pH 6.5) as an ion pair buffer and an elevated column temperature (60°C). Mobile phase composition was 70:30 (v:v) methanol, 28 mM TBAA, pH 6.5 (Solvent A) and methanol (Solvent B). A linear gradient from 5% B to 100% B over 44 min at a flow rate of 0.75 mL · min−1 was used. The method can provide quantitative results for up to 25 pigments with qualitative information for additional pigments. The 95% confidence limits were estimated as (a) 0.5%–3.8% for precision of replicate injections within and across sequences, (b) 3.2% for chl a calibration reproducibility, and (c) 5.1% for chl a method precision, including filter extraction and analysis.

Exponentially growing cultures of each strain were gently filtered onto a 25 mm GF/F glass fiber filter (Whatmann Inc., Florham Park, NJ, USA), flash-frozen on dry ice, and immediately stored at −80°C. The filters were subsequently transferred to 3.0 mL of acetone, sonicated for ∼60 s (Model 450; Branson Ultrasonic, Danbury, CT, USA) on ice, and filtered using 0.45 μm PTFE HPLC syringe cartride filters fitted with glass fiber prefilters. The extract was injected by mixing 150 μL of extract with 375 μL of 90:10, 28 mM aqueous TBAA solution (pH 6.5): methanol. The injection procedure combined a sample extract and buffer solution, respectively, in the 900 μL sample loop as follows: 150 μL buffer, 75 μL sample, 75 μL buffer, 75 μL sample, 150 μL buffer. The HPLC analyses were performed on an Agilent series 1100 HPLC at the Horn Point Analytical Services Laboratory.

Pigments were identified by retention times and absorption spectra identical to those of authentic standards and quantified against standards quantified at Horn Point Laboratory. Pigment standard was purchased in solution from DHI Water and Environment (Horsholm, Denmark). Chl a is purchased in solid form from Fluka (Millwaukee, WI, USA). To determine the mass spectrum of the two presumptive gyroxanthin diesters, the exact separation method performed at the Horn Point Analytical Services Laboratory was replicated at the Center of Marine Biotechnology on an Agilent 1100 Series LC/MSD with an APCI interface. The settings for the APCI interface were as follows: capillary voltage, 3,000 V; corona current, 4.0 uA; nebulizer pressure, 60 psi; drying gas flow, 5.0 L · min−1; drying gas temperature, 350°C; and vaporizer temperature, 475°C.

A factor analysis was performed on pigment:chl a ratios (by mass) of carotene, cis gyroxanthin diester, gyroxanthin diester, 19′ hexanoyloxyfucoxanthin, 19′ butanoyloxyfucoxanthin, chl c2, and chl c3 for all strains of K. veneficum and for a culture of K. mikimotoi. Factor analysis was performed in S-Plus 6.2 (TIBCO Software Inc., Palo Alto, CA, USA). Two factors were extracted from the data based on the results of a previously run principal components analysis. Kaiser varimax rotation was used.

Toxin identification.  Our criteria for detecting karlotoxins was operational. The karlotoxins should be released by filtration on glass fiber filters, removed from the filtrate by adsorption to C18 resins, and eluated by 60% to 80% methanol/water solutions. On a C8 reverse phase HPLC separation using a methanol to water gradient, karlotoxins in the methanol fractions were detected through their UV absorption at 225 nm (KmTx 1) and/or 235 nm (KmTx 2) with little UV absorption at 280 nm. Biologically active fractions were detected through hemolysis of red blood cells.

Toxin was obtained from different strains by growing two replicates of 1.0 × 108 cells (1–2 L of culture) for each strain, filtering onto 125 mm GF/F filters (Whatman). Toxin from the filtrate was concentrated with a 3 mL packed volume tC-18 solid phase extraction cartridge column (Waters Corp., Milford, MA, USA). After the filtrate was loaded onto the column, it was washed with 12 mL of increasing concentrations of methanol/water from 0% methanol to 80% methanol in 20% increments. The 80% methanol fraction was collected, dried under a vacuum, resuspended in 1 mL of methanol, and filtered with a GF/B filter prior to HPLC analysis.

Toxin samples were injected onto a C8 (LiChrosphere, 125 mm × 4 mm 5 micron bead size RP-8, Agilent) column and subjected to a 1 mL · min−1 10% to 95% methanol:water gradient over 25 min using an Agilent 1100 HPLC. Toxin peaks were detected using an Agilent Diode Array Detector (Model# G1315B) with a micro high-pressure flow cell (G1315B#020; 6 mm pathlength, 1.7 μΛ volume) over the wavelength range 190 to 350 nm. Based on the UV spectra, the absorption at 225 nm was used to detect KmTx 1, while absorption at 235 nm was used to detect KmTx 2. The entire UV spectra were saved for each UV detectable peak. The eluate from the DAD detector was split (1/3 to 1/6) using a graduated micro-splitter valve (Upchurch Scientific, Oak Harbor, WA, USA). The major portion of the eluate was fed to an Agilent 1100 fraction collector (Model G1364C), while the remaining portion was passed to the electro-spray nozzle of the MS (Agilent G1956A SL) for ionization with the following spray chamber conditions using nitrogen as the drying gas: flow rate 10 L · min−1, pressure 60 psi, temperature 350°C, fragmentor voltage 350 V, capillary voltage 4,000 V. A 1% (v/v) formic acid in water solution (Agilent Isocratic Pump at 0.1 mL · min−1) was added postcolumn via a T-connector to provide lower pH conditions for enhanced positive mode ionization. A 5 mM ammonium acetate solution in water was used for negative mode ionization. Fractions of 20 s duration (1/3 mL) during the first 32 min of the HPLC run were collected in 96-well plates to identify hemolytic peaks. For toxin mass determination, the hemolytic peaks were identified, and mass spectra were obtained for these peaks.

Peak areas from either the diode array detector or MS were used to quantify toxin quantities (Bachvaroff et al. 2008a). For the two CAWD isolates, there was no sufficient toxin for obtaining known weights of toxin, so interpolations from KmTx 1 and KmTx 2 data were used to estimate toxin amounts. To determine KmTx congener proportions, relevant mass ion abundances were obtained from the mass spectra and used in calculating percentages.

A 10 μL aliquot of each HPLC fraction was used in the hemolytic assay (Arzul et al. 1994) essentially as described by Deeds et al. (2002). Rainbow trout (Oncorhynchus mykiss) erythrocytes were used. Purified toxin dilutions were compared to standard curves of Saponin (Sigma-Aldrich, S-4521, St. Louis, MO, USA). For a 200 μL assay, it was observed that a linear response for Saponin occurred between 0.5 μg and 1.0 μg Saponin/well. Saponin standards and toxin samples were run in triplicate. Maximum hemolytic activity was calculated for each plate after 1 h incubation. Following the hemolytic assay, toxin concentration was estimated using at least two HPLC runs with appropriate extinction coefficients. For toxins derived from the two New Zealand isolates, extinction coefficients are unknown, so estimates of toxin amount were used based on the similarity of the toxin to KmTx 1 or KmTx 2 as described below. The dose response curves [Fractional Hemolysis vs. log (μg · mL−1) of karlotoxin] were fitted to the Hill equation:


using a nonlinear regression (Igor 6.02, Wavemetrics, Lake Oswego, OR, USA). In all cases, convergence to defined parameters was observed.


Species identity.  Each athecate gymnodinoid strain had two to four chloroplasts with a straight apical groove and a ventral pore (Fig. 1A inset for CCMP 1974). The epicone was conical or rounded, while the hypocone was always hemispherical. The cingulum was left-handed and could be displaced as much as two cingulum widths. The sulcus extended into the epicone. The cells were typically golden-brown in color. The nucleus was large and rounded, usually situated centrally in the cell or slightly to left side of the hypocone. The range in cell diameter (Table 1) was from 8.29 to 12.07 μm with a median value of 10.30 μm. In some strains, a feeding appendage (peduncle or tow line) could be observed when prey (i.e., cryptophytes) were provided. Some of the strains (e.g., F4) frequently exhibited bimodal size distributions, and in some cases, planozygotes (two trailing flagella) could be observed in the culture. Cysts were never observed in any of the cultures.

Figure 1.

 Location of origin for Karlodinium veneficum cultures used in this study. (A) bsl00001 = KmTx 1 producer, • = KmTx 2 producer; (B) • = CAWD 83. Inset: Photomicrography of CCMP 1974 showing the ventral pore (VP) and apical groove (AG). The site of origin of CAWD 66 is unknown. Multiple strain isolations from a single location are indicated by a number above the symbol. (C) A parsimony tree based on ITS sequences from the 18 K. veneficum strains in this paper. Note that there are no sequence differences between the 16 strains from the U.S. Atlantic coast and the CAWD66 isolate from New Zealand over 496 bp.

Of the 18 strains examined, 17 had identical ITS sequences over the 492 bases examined (Fig. 1C). The CAWD83 culture had a single point mutation and a deletion site when compared to the other strains. All strains produced positive results in the quantitative PCR assay for the plastid 16S SSU gene with a mean CT of 21.07 ± 2.4 (n = 18) for a 20 ng DNA input, suggesting identical plastid SSU DNA sequences for all strains (Table 1).

Pigments.  As seen in Figure 2 and Table 2, compared to the peridinin containing dinoflagellate Amphidinium carterae (Fig. 2A), all 18 strains of K. veneficum exhibited a complete absence of peridinin with the presence of fucoxanthin, 19′ hexanoyloxyfucoxanthin, 19′ butanoyloxyfucoxanthin, and gyroxanthin diester pigments (Johnsen and Sakshaug 1993), similar to its sister genera Karenia (K. mikimotoi, Fig. 2D). Different from Karenia were the nearly equivalent quantities of the two gyroxanthin diesters in all North American strains of K. veneficum. The UV spectra were identical for both peaks, and the LC–MS analysis using APCI ionization found an identical mass (m/z 866) for both peaks. We therefore attribute these two peaks to the 9-cis and all trans gyroxanthin diesters isolated from K. veneficum (Gymnodinium galatheanum) and characterized by Bjørnland et al. (2000).

Figure 2.

 HPLC pigment analysis of (A) the peridinin-pigmented Amphidinium carterae (CCMP 1314), compared to (B) and (C) the 19′hexanoyl-fucoxanthin pigmented Karlodinium veneficum (two strains), and (D) Karenia mikimotoi (CCMP429). Note the nearly equivalent quantity of the gyroxanthin diesters in the North American strains of K. veneficum compared to K. mikimotoi. Pigment designations are: Chl c3 (chlorophyll c3), Chl c2 (chlorophyll c2), Chl c1 (chlorophyll c1), Per (peridinin), But (19′-butanoyloxyfucoxanthin), Fuco (fucoxanthin), Hex (19′-hexanoyloxyfucoxanthin), Diadino (diadinoxanthin), Zea (zeaxanthin), Chl a (chlorophyll a), β-Car (β-carotene).

Table 2. Karlodinium veneficum strain pigments.
Culture nameChl c3/ Chl a (ng · μg−1)Chl c2/ Chl a (ng · μg−1)Chlorophyllide a/Chl a (ng · μg−1)Phaeophorbide a/ Chl a (ng · μg−1)19′Butanoyloxy Fucoxanthin/ Chl a (ng · μg−1) Fucoxanthin/ Chl a (ng · μg−1) Violaxanthin/ Chl a (ng · μg−1) 19′Hexanoyloxy Fucoxanthin/ Chl a (ng · μg−1) Diadinoxanthin/ Chl a (ng · μg−1) Diatoxanthin/ Chl a (ng · μg−1) Zeanxanthin/ Chl a (ng · μg−1)Gyroxanthin diester/ Chl a (ng · μg−1)cis Gyroxanthin diester/Chl a (ng · μg−1)Monovinyl Chl a/Chl a (ug · μg−1)Phaeophytin a/ Chl a (ng · μg−1) Carotenes/ Chl a (ng · μg−1)Chl a (pg · cell−1)
  1. Pigments not detected: chl c1, perdinin, neoxanthin, prasinoxanthin, anteraxanthin, alloxanthin, myxoxanthophyll, lutein, canthaxanthin, chl b, divinyl chl a.

CCMP 197572.8139.
CCMP 197451.6137.
CCMP 238877.0129.
CCMP 228260.3139.
CCMP 228367.7150.
CCMP 206474.6148.03.10.0168.3124.20.00327.9124.759.420.
CCMP 277880.0132.82.91.1370.7312.34.1134.750.029.40.0048.229.90.99420.816.9 
K. mikimotoi86.5118.

When we performed a factor analysis of the pigment profiles for the 18 strains, a clear separation between the North American strains and New Zealand strains was evident (Fig. 3). Factor 1 accounted for 39% of the variance in the pigment ratios data and was loaded with (+)19′ hexanoyloxyfucoxanthin, (+)19′ butanoyloxyfucoxanthin, (+) gyroxanthin diester, (+) chlorophyll c2, and (−) fucoxanthin (Fig. 3A). Factor 2 accounted for an additional 19% of the variance in the pigment ratios data set and was loaded with (+) chlorophyll c3, (+) carotene, and (−) cis gyroxanthin diester (Fig. 3A). The biplot of factor 1 versus factor 2 scores (Fig. 3B) shows a tight cluster along the factor 1 axes and contains all K. veneficum strains except CCMP 2064, CAWD 66, and CAWD 83, indicating that K. veneficum strains from the eastern seaboard of the U.S. (excluding strain CCMP 2064) show similarities among their fucoxanthin:chl a ratios and their cis gyroxanthin diester:chl a values. CCMP 2064 was separate from this group due to its factor 1 score, which was more similar to the factor 1 score observed for CAWD 66 and CAWD 83, indicating that CCMP 2064 differed from the eastern seaboard of the U.S. cluster due to higher 19′ hexanoyloxy fucoxanthin:chl a and 19′ butanoyloxyfucoxanthin:chl a ratios, lower fucoxanthin:chl a ratios than other U.S. eastern seaboard strains, but similar cis gyroxanthin:chl a.

Figure 3.

 Factor analysis of accessory pigment:chl a ratios for strains of K. veneficum and an outgroup, K. mikimotoi. (A) Factor loadings, indicating two axes of variability within the pigment ratios data set. Points grouped along either the x- or y-axis are positively correlated to each other, while points opposite each other along either axis are negatively correlated. (B) Factor scores biplot. Clustered points are similar with regard to the attributes represented by factors 1 and 2, while separated points differ.

K. mikimotoi was segregated from the rest of the data by both lower factor 1 and higher factor 2 scores, indicating a low cis-gyroxanthin diester:chl a similar to the CAWD strains and higher fucoxanthin:chl a than any of the K. veneficum strains analyzed. CAWD 83 and CAWD 66 formed a distinct group, separated from the others based on higher scores on both axes, standing out as having lower cis-gyroxanthin:chl a compared to the eastern U.S. seaboard strains and also as having higher 19′ hexanoyloxy fucoxanthin:chl a and 19′ butanoyloxy fucoxanthin:chl a ratio, and lower fucoxanthin:chl a than U.S. eastern seaboard strains. Overall, based on the cell morphology, molecular ribosomal sequence signatures, and the pigment profiles, all 18 strains are authentic K. veneficum.

Toxin phenotype.  Our current understanding of karlotoxins indicates that they occur in two basic forms (Bachvaroff et al. 2008a): KmTx 1 and KmTx 2. The KmTx 1 toxins have a shorter UV absorption maximum than KmTx 2 toxins (absorption maximum: 225 nm vs. 235 nm) and do not contain chlorine. Within these two groups, various hydroxlated congeners can be detected as well as various chain length variants (Bachvaroff et al. 2008a). In these HPLC runs, an additional later-eluting peak was frequently observed when growing cultures in natural seawater. This peak produces multiple MS ions differing by 44 amu and is likely to be a plasticizer polymer contaminant, which shows hemolytic activity.

KmTx 1 phenotypes.  Of the six strains examined from the Chesapeake, one strain, MD5, produced no measurable toxin (Fig. 4A). Two other strains isolated at the same time and location, MD2 and MD6, produced only a single hemolytic peak identified as KmTx 1-3 (MS 1345, M+Na) based on elution time, hemolytic activity, and UV spectrum (Fig. 4B). The other two KmTx 1–producing strains, CCMP 1974 and 1975, exhibit two major hemolytic toxins, KmTx 1-1 (MS 1331, M+Na) and KmTx 1-3 (Fig. 4C). These toxins have mass spectra that differ by 14 amu (Bachvaroff et al. 2008a). There is a minor additional KmTx 1 form that eluates between KmTx 1-1 and KmTx 1-3, which we call KmTx 1-2 (MS 1381.9, M+Na).

Figure 4.

 Liquid chromatography traces for partially purified toxin extracts demonstrating activity and elution time for toxin types found in U.S. Atlantic coast strains. The large dot indicates a mixture of plasticizers that coelutes from a C-18 column with karlotoxin. The absorbance values (225 nm for KmTx 1 or 235 nm for KmTx 2) are shown as a black line (left axis) with the hemolytic activity of 20 s fractions overlaid (right axis). (A) The nontoxic MD5 strain. (B) Strains containing KmTx 1-3. (C) Strains containing KmTx 1-1 and 1-3. (D) Strains containing KmTx 2 with a UV absorbance maximum at 235 nm.

KmTx 2 phenotypes.  The strains from south of the Chesapeake Bay all contained KmTx 2 (MS 1344, M+Na) as the major karlotoxin based on elution time, hemolytic activity, and UV spectrum (Fig. 4D). We also observed in some strains (e.g., CCMP 2283) a minor peak preceding the major KmTx 2 peak with a mass of 1307.8 (1330.8, M+Na). The isotope distribution of the mass spectra for this peak was consistent with the absence of a chlorine.

Both KmTx 1 and KmTx 2 producing strains also exhibited minor eluting peaks earlier in the chromatograms (at ∼15 min), which appear to be sulphated forms of the karlotoxins (Bachvaroff et al. 2008a). These compounds are larger than the parent karlotoxins by 102 amus and appear to be sulfated forms of KmTx 1-1, KmTx 1-3, and KmTx 2. This sulfated form of karlotoxin appears to be the same as the ToxB karlotoxin described by Deeds et al. (2002). The sulfated form is readily detected in culture filtrates when using the filter adsorption technique (Bachvaroff et al. 2008a).

Karlotoxin hydroxylated congeners.  Each chromatographic peak assigned to be a karlotoxin based on hemolytic activity (e.g., Figs. 3 and 4) was determined to have two dominant mass ions that differ by 16 amu in their mass spectra (Bachvaroff et al. 2008a). While the ratio of these congeners for KmTx 1-3 is the same in both MD2 and MD6 (Fig. 5A), there are differences in the congener ratios for KmTx 1-1 between CCMP 1974 and 1975 (Fig. 5, B and C). CCMP 1975 has a larger proportion of the smaller 1331.8 ion to the 1347.8 ion when grown under the same conditions. Other ions that do not differ by 16 are also present in several spectra, but it remains unclear if these are genuine minor congeners or are artifacts. There were also some other subtle differences in MS profile between the strains. Specifically, the F4 strain (Fig. 5D) had a larger proportion of ions 16 amu smaller (1351.8) when compared with the other strains (Fig. 5E) where the ion distribution was shifted to a gain of 16 amu (1383.8). Recent structural characterization indicates that these congeners represent different states of hydroxlation of the base karlotoxin molecule.

Figure 5.

 Mass spectra for the peaks with hemolytic activity. All spectra are positive mode and are likely to be sodium adducts. (A) KmTx 1-3 mass spectrum typical of both MD2 and MD6 strains as well as the KmTx 1-1 peak from CCMP1974 and 1975. (B) The KmTx 1-1 mass spectrum from CCMP1975 has a different congener ratio when compared to the same peak from CCMP1974 (C). (D) Shows the mass spectrum for KmTx 2 from strain F4 with a different congener ratio when compared to the mass spectrum for the other KmTx 2 strains (E).

New Zealand strains.  The two strains from New Zealand produce different hemolytic toxins, with the CAWD83 toxin having a 225 nm absorption maximum (KmTx 1-like) and a molecular weight of 1287.8 (M+Na), and the CAWD66 toxin having a 235 nm absorption maximum (KmTx 2-like) and a molecular weight of 1291.8 (M+Na). Both toxins elute earlier from the C8 column than the karlotoxins detected in the North American strains (i.e., KmTx 1 and 2) (Fig. 6A). As with the North American karlotoxins, the toxins from the New Zealand strains coelute as hydroxylated congeners containing mass ions that are 16 amu larger, as well as other less abundant ions (Fig. 6, B and C).

Figure 6.

 (A) Liquid chromatography traces for partially purified C-18 extracts from the two New Zealand strains, CAWD83 and CAWD66. The large dot indicates a mixture of plastic esters that coelutes with karlotoxin on the C-18 column. Hemolytic activity is included on the right axis for 20 s fractions. (B and C) Mass spectra for the active peaks from these two strains, which also include 16 amu larger congeners. (D) UV spectra for the two different toxins.

There were insufficient pure toxins to determine response factors for quantification in the New Zealand strains, so quantities were estimated based on the extinction coefficients of KmTx 1-1 and 1-3 (694 ± 8.29 and 633 ± 6.51 mAU s−1 · μg−1, respectively; Bachvaroff et al. 2008a) for CAWD83, which shares an absorption maximum of 225 nm with these toxins. For the CAWD66 toxin, the KmTx 2 extinction coefficient (478 ± 3.23 mAU s−1 · μg−1; Bachvaroff et al. 2008a) was used because the toxin had an absorption maximum of 235 nm (Fig. 6D).

Growth parameters under batch growth.  Under these standardized conditions (100 μm photon · m−2 · s−1 PAR [14:10 L:D], 20°C, 15 psu), most of the strains reached a maximum density of >100,000 cells · mL−1 during the 25 d growth cycle (Table 1). Only F4, FB3, and CCMP 2778 did not reach this density (see Table 1). Several strains (MD2, MD5, Slocum, IB4, CCMP 2282, PD-6, and CAWD83) exceeded 200,000 cells · mL−1. Cell division rates based on these data ranged from 0.17 to 0.36 division · d−1 and were statistically identical (Mean: 0.26 ± 0.06 division · d−1). The ANOVA performed on division rate (F19,40 = 2.7, P = 0.006), maximum density (F19,40 = 60.3, P < 0.001), and toxin cell quota (F18,37 = 16.1, P < 0.001) indicated significant differences among strains.

Toxin cell quotas during batch growth. KmTx 1 phenotypes:  Karlotoxin was detected at all stages of growth for those strains producing toxin. MD5 never produced toxin throughout its growth curve (data not shown). The KmTx 1–producing cultures have in general less toxin per cell than the KmTx 2 strains, with values ranging from 0 to 0.252 pg · cell−1 (Mean: 0.09 ± 0.12 pg · cell−1). Overall, the cell quota of toxin was stable for most strains throughout the growth cycle (Fig. 7) with the exception of MD2, which had a notable increase in toxin cell quota after reaching peak density. For strains producing both KmTx 1-1 and KmTx 1-3, the ratio between toxins remained reasonably constant throughout the growth curve (Fig. 7; CCMP 1974 and CCMP 1975).

Figure 7.

 Cell density (1,000 × cells · mL−1; left axis) and karlotoxin cell quota (pg · cell−1; right axis) for KmTx 1 producing strains during the course of a 25–30 d batch culture growth curve.

KmTx 2 phenotypes:  The toxin cell quota for KmTx 2 ranged from 0.037 to 2.99 pg · cell−1 (mean 1.46 ± 1.29 pg · cell−1) between strains (Table 1). Generally, the toxin cell quota was stable during the course of growth, although two strains (CCMP 2064 and MBM1) exhibited decreasing amounts in the later stages of growth (Fig. 8). When normalized to biovolume changes (e.g., CCMP 2778) during the growth curve (Fig. 9), toxin cellular concentrations remained much more stable than when normalized per cell.

Figure 8.

 Cell density (1,000 × cells · mL−1; left axis) and karlotoxin cell quota (pg · cell−1; right axis) for KmTx 2 strains during a 25–30 d batch culture growth curve.

Figure 9.

 KmTx 2 cell quotas normalized to biovolume for CCMP 2778 during a 25–30 d batch culture growth curve.

New Zealand strains:  The relative amounts of CAWD83 toxins in the New Zealand strain showed an almost constant increase in the amount of toxin per cell over the full course of the culturing cycle (Fig. 10); for CAWD66, the toxin cell quota was constant until after the cell numbers had decreased when toxin cell quota increased.

Figure 10.

 Cell density (1,000 × cells · mL−1; left axis) and karlotoxin cell quota (pg · cell−1; right axis) for CAWD 88 and CAWD 66 strains during a 25–30 d batch culture growth curve.

Karlotoxin potency.  The potency of purified karlotoxins is presented in Figure 11. In all cases, the purity of each karlotoxin exceeded 90%, except for KmTx 1-2, which was only 45% pure, the remaining contaminates being equal amounts of KmTx 1-1 and KmTx 1-3. There was nearly a 500-fold difference in hemolytic potency (HD50%) between the most active (KmTx 1-2 mixture) and least active (CAWD66) karlotoxin (Table 3). The overall ranking was KmTx 1-2 mixture > KmTx 1-3 > = 1-1 > CAWD83 > KmTx 2 > CAWD66 toxin > Saponin (on a per-weight basis, Table 3). Complete hemolysis occurred over less than a factor of 10 increase in karlotoxin concentration.

Figure 11.

 Comparative hemolytic activity for the different karlotoxins using rainbow trout erythrocytes compared to saponin. Note the logarithmic scale for toxin amount. The fitted line is based on the Hill equation, and the HD50% estimates for these curves are presented in Table 3.

Table 3.   Summary of masses and hemolytic activities of karlotoxins examined.
Toxin typeMassaHD50% (μg · mL−1)Strains
  1. aOn LC–MS these will appear as sodiated ions (i.e., +23).

KmTx 1-11308.80.192 ± 0.007CCMP 1974
 Hydroxy-KmTx 1-11324.8 CCMP 1975
KmTx 1-2 Mixture1358.80.045 ± 0.021 
KmTx 1-31322.80.13 ± 0.003 
 Hydroxy-KmTx 1-31338.8 MD2 and MD6
KmTx 21344.81.27 ± 0.05 
 Hydroxy-KmTx 21360.8 All other KmTx 2 strains
 Dehydroxy-KmTx 21328.8 F4
CAWD831268.9 & 1284.81.41 ± 0.46 
CAWD661280.8 & 1264.813.75 ± 0.001 
Saponin 3.17 ± 0.22 

Strain cellular potency.  Using the karlotoxin cell quotas presented in Table 1 for each strain and the HD50% estimates given in Table 3, we calculated an equivalent saponin potency per cell for each strain by dividing the HD50% for saponin by the HD50% for each karlotoxin and then multiplying this value times the karlotoxin cell quota for each strain. The results of this exercise are presented in Figure 12, where we plot toxin cell quota versus saponin cell equivalents on a log–log relationship. Clearly, the difference between KmTx 1–producing strains and KmTx 2–producing strains is readily apparent, as is the large variation in toxin cell quotas among the strains examined. Note also that one of the New Zealand strains (CADW83) falls on the same trend line as the KmTx 2 strains, while the other New Zealand strain (CADW66) is an outlier.

Figure 12.

 Log–Log scatter plot of karlotoxin cell quota versus saponin picoequivalents per cell. The solid symbols represent the KmTx 2 strains, and the open symbols the KmTx 1 strains presented in Table 1. The star symbols represent the two New Zealand strains.


Karlotoxin plays several important roles in the biology of K. veneficum. It is the agent responsible for fish kills due to K. veneficum (Deeds et al. 2002, 2006, Kempton et al. 2002) and obeys a dose-response relationship for fish mortality (Deeds et al. 2006). Susceptibility to karlotoxin depends on the membrane sterol composition of target organisms (Deeds and Place 2006). Species with a preponderance of 4α-methyl sterols (e.g., gymnodinosterol, dinosterol, or amphisterol) appear immune to karlotoxin, while those having desmethyl sterols (e.g., ergosterol, cholesterol, or brassicasterol) show reduced growth, inhibited feeding, or death in the presence of karlotoxin (Adolf et al. 2006, 2008a, Deeds and Place 2006). Low-level toxic or nontoxic strains are unlikely to cause fish kills at densities typically occurring in nature. Fish kills are a by-product of K. veneficum toxin. The division rate of K. veneficum is 2- to 3-fold faster as a mixotroph that ingests algal prey than as a phototroph (Li et al. 1999), with toxin playing a role in prey capture (Adolf et al. 2006, 2008b). While high in situ grazing rates on K. veneficum have been observed (Johnson et al. 2003), experimental comparisons of toxic versus nontoxic K. veneficum as prey support the conclusion that KmTx can deter grazing by some organisms (Adolf et al. 2007, 2008b, Brownlee et al. 2008, Stoecker et al. 2008, Waggett et al. 2008). Toxic strains will likely outcompete nontoxic strains in mixed assemblages where prey and predators are both present, but a potential drawback of high cellular toxicity is increased mortality due to the parasite Amoebophrya sp. (Bai et al. 2007). Given the key role that toxin plays in the ecology of K. veneficum, we sought to examine toxin production in detail from a diverse sample of strains. Our data showing a wide range of cellular toxicity between strains, particularly strains originally isolated from the same water sample (MD2, MD5, MD6), suggest that K. veneficum strains of different toxicity exist in any given natural population.

Morphology, photopigments, and ITS ribotype of K. veneficum strains.  As a first step in characterizing this strain collection, we wanted to identify each culture as K. veneficum using species level features including ITS sequence, a TaqMan assay for plastid ribosomal RNA, and LM. The variation in toxin type and amount between K. veneficum strains, consistent with our previous studies using strains only from the U.S. eastern seaboard (Deeds et al. 2004), contrasts sharply with their homogeneity using morphology, pigment composition, and ITS ribotype. All 18 strains exhibited the characters that define the genus Karlodinum using LM (Daugbjerg et al. 2000, Bergholtz et al. 2005). Although the full genetic characterization of K. veneficum strains is the subject of future study, in our current study, toxin phenotype is less conserved than ITS sequence, a commonly used marker for species and/or subspecies designation.

One of the synapomorphic characters that distinguish the Kareniaceae (i.e., Karenia, Takayama, and Karlodinium) relates to the presence of fucoxanthin and its derivatives. As a taxonomic marker, multivariate analysis of accessory pigment:chl a ratios produced similar results as seen in the analysis of the ITS sequence. Both ITS sequence and pigment profiles produced a distinct group consisting of U.S. eastern seaboard strains (although CCMP 2064 grouped separately in pigments) and, unlike the ITS sequence, clearly separated both the New Zealand strains from the U.S. strains. Neither character had sufficient resolution to distinguish within the group of U.S. strains. From a physiological perspective, our photopigment results are consistent with those obtained by Stolte et al. (2000) for Emiliania huxleyi, showing that the ratio of total fucoxanthin:chl a is constant, while the ratios of each fucoxanthin may vary either within or between strains. An unexpected pigment to find for K. veneficum was zeanxanthin, which was observed in trace amounts with MD2, MD5, MD6, and CCMP 2064. We do not believe that this is an intrinsic pigment for K. veneficum but represents a cyanobacterial contamination in the culture.

Toxin production during batch culturing.  In this study, no attempt was made to determine the optimal growth conditions for each strain, but rather to compare them under uniform laboratory conditions, so these data do not necessarily define the maximum toxin production for each strain. There were no statistically significant differences in phototrophic growth rates for all 18 strains (0.26 ± 0.06 μ; Table 1), similar to observations with P. micans (Costas 1990). This finding suggests that interstrain differences in toxin production were not due to differences in growth rate and implies that there is a significant metabolic cost associated with toxin production. However, maximum densities were significantly different with a range of ∼35,000 cells · mL−1 to >260,000 cells · mL−1. We believe that some of this difference can be attributed to the lower salinity (15 psu) used since strains like CCMP 2778 can attain densities >200,000 cells · mL−1 when grown at higher salinities (e.g., 30 psu). The differences in maximal cell density could also reflect differences between strains in pH tolerance since these are batch cultures. When we attempted to correlate cell size and cell yield with toxin per cell amounts, we found no significant correlation.

We observed a generally stable or decreasing toxin cell quota as cells moved from exponential to stationary or senescent growth phase. The exceptions were CCMP 2064 and MBM1, which produced less toxin per cell in stationary and declining phases of culture, and MD2, which produced more toxin. In f/2 medium, pH and inorganic carbon limitation are more likely to limit growth than macronutrients (N or P) (Schmidt and Hansen 2001), so the relatively constant level of KmTx per cell we observed in most strains over the growth cycle likely reflects simultaneous pH or inorganic carbon limitation of cellular growth and toxin synthesis due to the carbon-rich nature of karlotoxin (Bachvaroff et al. 2008a). We note, however, that this trend contrasts with the loss of toxicity in Chrysochromulina polylepis observed in high pH stationary phase conditions (Schmidt and Hansen 2001).

Karlotoxin phenotype.  Karlotoxin type (i.e., KmTx 1 vs. KmTx 2) is genetically determined since no change in toxin type was found during the course of the growth cycle. Our findings confirm the geographic and genetic distinction between KmTx 1 and KmTx 2 strains along the U.S. eastern seaboard (Deeds et al. 2004, Bachvaroff et al. 2008b), but the two novel toxins from the New Zealand cultures have earlier elution times and lower masses than the previously described KmTx 1 and KmTx 2. Based on UV spectra, the toxin from CAWD83 was assigned to the KmTx 1 group, and the other New Zealand isolate, CAWD66, belongs to the KmTx 2 group (Fig. 6D). While we appreciate the somewhat arbitrary use of UV spectra to distinguish between KmTx 1 and KmTx 2, this difference reflects different double-bond conjugations underlying one of the chemical differences between the karlotoxins.

The nontoxic strain MD5 was isolated at a location in the Chesapeake Bay where a number of toxic strains were also derived (Fig. 4, Table 1). The occurrence of nontoxic strains is not unexpected for this species. We suggest that Gymnodinium vitiligo, originally isolated as a morphologically identical but nontoxic species related to Gymnodinium veneficum strain PLY 103 (syn. K. veneficum) (Ballantine 1956), was also a nontoxic strain of K. veneficum.

As was previously shown (Bachvaroff et al. 2008a), the individual HPLC peaks on a C18 column contain multiple hydroxylated congeners (Fig. 5). In most cases, these congeners differ by 16 amu so that for KmTx 2, for example, there are a series of three congeners starting from a mass of 1351.8 amu and stepping up in 16 amu increments to 1383.8 amu, most likely reflecting the addition of a single oxygen atom. These apparent congeners are separable on a normal phase amine column, have distinct MSn spectra and are not present when using the same HPLC–MS system with amphidinols (T. R. Bachvaroff and A. R. Place, unpublished data). Therefore, it is likely that these compounds are not artifacts of the isolation and detection process, but rather represent a pool of slightly differently hydroxylated toxins in the cell. Fine-scale differences within KmTx 1 (Fig. 5, B vs. C) or KmTx 2 (Fig. 5, D vs. E) strains further divided these populations and were stable throughout the growth cycle. A pattern of multiple congeners was also observed, with CAWD83 consisting of two pairs of compounds with a mass difference of 16 amu each and a mass difference of 58 amu between the two groups. These results imply that genetically diverse K. veneficum populations have subtle diversity in their toxin profiles.

We noted with time that some strains changed their toxin phenotype (cell quota and type of karlotoxin produced) as we grew them phototrophically in large quantities for many generations in replete media. The general trend was to lower karlotoxin cell quotas with production of new karlotoxins not observed initially. The change appeared to be permanent since single-cell reisolation did not return the original phenotype. For example, CCMP 2064 went from largely (96.9%) producing KmTx 2 (1344.8 amu) to largely (79%) producing a later eluting congener (1322.8 amu). When CCMP 2778 was cultured in large volumes (40 L) for 6 weeks with six 40 L harvests, its toxin phenotype changed from producing KmTx 2 exclusively to producing KmTx 2 and a new earlier eluting congener (1310.9 amu) at 22% of the total karlotoxin. Similarly, CCMP 1974 stopped producing KmTx 1-3 and produced a much earlier eluting congener (1310.9 amu) along with KmTx 1-1. The reason for these changes is unclear at this time.

Strain cellular potency.  Accounting for the differences in activity between the strains, the activity per cell was calculated using hemolytic equivalents (Fig. 12). Interestingly, the higher potency of KmTx 1 compounds is correlated with a lower cell quota; hence, KmTx 1 and 2 containing strains have similar ranges of cellular toxic activity. Aside from the nontoxic MD5 strain, CCMP 2388 and CAWD66 have very low toxic activity.

The diversity in toxin production between different cultured K. veneficum strains likely reflects specific selection parameters in the environment. Selection for toxin-producing strains that are less palatable as food and more successful as predators (Adolf et al. 2007) has to be balanced by the potential for increased parasitism by Amoebophyra (Bai et al. 2007), at least under conditions where parasites have significant impacts on the K. veneficum population. Furthermore, toxin production might be tailored to impact specific predators or prey in different environments, helping to explain the large diversity in toxin profiles between strains observed here.

Photopigment analyses and molecular phylogenies based on nuclear rDNA sequences place the fucoxanthin-containing dinoflagellates K. veneficum and Karenia sp. near each other (Daugbjerg et al. 2000, Guillo et al. 2002, Bergholtz et al. 2005), but distant from the peridinin-containing Amphidinium sp. (Daugbjerg et al. 2000). In terms of toxin, however, the linear polyketides produced by K. veneficum and Amphidinium carterae are very similar to each other in structure and bioactivity (Bachvaroff et al. 2008a), and both are very different from the ladder polyketides produced by members of the Karenia genus. One possible cause of this similarity is the horizontal transmission of toxin genes between Amphidinium and Karlodinium. Current estimates for the minimum number of malonate extensions involved in amphidinol synthesis is ∼30, which would translate into nearly a 150 kb operon that would have to be transferred. There are currently no data for either species, which would indicate that such an extrachromosomal element exists.

Conclusion.  The results of this study suggest that toxin production in K. veneficum strains is genetically determined, present throughout the culturing cycle, and the toxin cell quota is highly variable. Toxin production ranges from nondetectable to the range of picograms per cell, with the majority of strains showing some toxin production. These results also suggest that under certain conditions of growth, different strains will respond differently, either increasing or decreasing toxin production, but that for the most part, toxin per cell quota is stable during the growth cycle. These results further highlight the problems of working with a diverse and largely uncharacterized species such as K. veneficum. While this study certainly represents an advance in terms of number of strains and toxin types established, it is clear that substantial diversity remains undiscovered in this globally distributed harmful algal species.


This paper is partially a result of research funded by the National Oceanic and Atmospheric Administration Coastal Ocean Program under award #NA04NOS4780276 to University of Maryland Biotechnology Institute and Grant # U50/CCU 323376, Centers for Disease Control and Prevention and the Maryland Department of Health and Mental Hygiene. This is contribution # 08-187 from the Center of Marine Biotechnology and contribution # 276 from the ECOHAB program.