Bacterial Community Affects Toxin Production by Gymnodinium catenatum

The paralytic shellfish toxin (PST)-producing dinoflagellate Gymnodinium catenatum grows in association with a complex marine bacterial community that is both essential for growth and can alter culture growth dynamics. Using a bacterial community replacement approach, we examined the intracellular PST content, production rate, and profile of G. catenatum cultures grown with bacterial communities of differing complexity and composition. Clonal offspring were established from surface-sterilized resting cysts (produced by sexual crosses of strain GCDE06 and strain GCLV01) and grown with: 1) complex bacterial communities derived from each of the two parent cultures; 2) simplified bacterial communities composed of the G. catenatum-associated bacteria Marinobacter sp. strain DG879 or Alcanivorax sp. strain DG881; 3) a complex bacterial community associated with an untreated, unsterilized sexual cross of the parents. Toxin content (STX-equivalent per cell) of clonal offspring (134–197 fmol STX cell−1) was similar to the parent cultures (169–206 fmol STX cell−1), however cultures grown with single bacterial types contained less toxin (134–146 fmol STX cell−1) than offspring or parent cultures grown with more complex mixed bacterial communities (152–176 fmol STX cell−1). Specific toxin production rate (fmol STX day−1) was strongly correlated with culture growth rate. Net toxin production rate (fmol STX cell−1 day−1) did not differ among treatments, however, mean net toxin production rate of offspring was 8-fold lower than the parent cultures, suggesting that completion of the sexual lifecycle in laboratory cultures leads to reduced toxin production. The PST profiles of offspring cultures were most similar to parent GCDE06 with the exception of cultures grown with Marinobacter sp. DG879 which produced higher proportions of dcGTX2+3 and GC1+2, and lower proportions of C1+2 and C3+4. Our data demonstrate that the bacterial community can alter intracellular STX production of dinoflagellates. In G. catenatum the mechanism appears likely to be due to bacterial effects on dinoflagellate physiology rather than bacterial biotransformation of PST toxins.

Putative biosynthetic genes and pathways for PST production are now known from several cyanobacteria [7], [8], [9], and homologous genes have been detected from PST-producing dinoflagellates including G. catenatum [10]. There is currently no convincing structural or molecular evidence that heterotrophic bacteria produce PST autonomously yet the bacterial community is known to indirectly influence dinoflagellate PST toxicity [11], [12], [13] either by biotransformation from one PST derivative to another [14], [15], [16] or potentially via their effect on dinoflagellate growth and physiology [17], [18]. For example, axenic Alexandrium and Protogonyaulax cultures have been shown to contain a higher PST content than their non-axenic counterparts, suggesting that the associated bacterial community may reduce dinoflagellate toxin production under some circumstances [18], [19].
Laboratory-grown G. catenatum cultures are associated with a suite of associated bacteria [11], [24], in which the twoproteobacteria, Marinobacter sp. and Alcanivorax sp., are constant components [11], [24]. Using strains originally isolated from the Tasmanian G. catenatum strain GCDE08 [11], Bolch et al., [25] found that G. catenatum has an obligate requirement for bacterial associates and that both Marinobacter sp. DG879 and Alcanivorax sp. DG881 are capable of supporting survival and growth of the vegetative G. catenatum cells. Given this high reliance on bacterial associates, we anticipated that the bacterial community could alter toxin content and/or production of the dinoflagellate cell.
Our earlier studies have found no consistent link between community membership and cellular toxicity [11]; however, these studies did not examine PST production rates and were potentially confounded by isolation, geographic and culture-related effects.
Here we use controlled uni-bacterial community replacement [25] to show that changes to the associated bacterial community modify both PST content and production by the dinoflagellate G. catenatum.

Production of resting cysts
Two sexually compatible, PST-producing, Tasmanian strains, GCDE06 (CS-301/06) and GCLV01 (CS-800), were provided by the Australian National Algae Culture Collection (ANACC; http://www.csiro.au/ANACC). Sexual crosses were carried out in sterile 55 mm diameter plastic Petri dishes containing 10 mL of nitrate and phosphate deficient GSe medium [26] and incubated at 21uC62uC, at 80 mmol PAR m 22 sec 21 with a 18:6 h light:dark cycle for 3 weeks to promote cyst formation. Resting cysts were harvested by manual micropipetting, washed several times in fresh GSe medium, surface-sterilized in 0.5% H 2 O 2, and washed 3 times to remove residual H 2 O 2 [25]. Successful surfacesterilization of cysts was checked by spread-plating 10 mL of the sterilized sample onto ZM1 agar incubated at 24uC for 4 days. If bacterial growth was observed after incubation, the cysts were discarded.

Unibacterial and community replacement cultures
A community replacement approach [25] was used to establish a series of clonal offspring cultures with the following modified bacterial communities: 1) complex bacterial communities derived from each of the two parent cultures GCDE06 and GCLV01; 2) simplified uni-bacterial communities of the G. catenatum-associated bacteria Marinobacter sp. DG879 or Alcanivorax sp. DG881; or 3) a complex bacterial community derived from unsterilized cysts from a sexual cross of both parent cultures. Table 1 summarises all cultures established in this study. Bacterial strains Marinobacter sp. DG879 and Alcanivorax sp. DG881 were originally isolated from the G. catenatum strain GCDE08 [24] and maintained on modified ZoBell's Marine agar (ZM1) [25], [27] at 20uC in total darkness. Groups of ten to fifteen sterilized G. catenatum cysts were each placed into 55 mm diameter sterile polystyrene Petri dishes containing 10 mL GSe medium using a micropipette. A specific bacterial community was then added to each of the Petri dishes, with triplicate Petri dishes established for each of the four different bacterial communities as follows: Community 1) 1 mL of 10 5 cells mL 21 of Marinobacter sp. DG879; Community 2) 1 mL of 10 5 cells mL 21 of Alcanivorax sp. DG881; Community 3) 1 mL of 10 5 cells mL 21 of a 5 mm filtrate from the non-axenic parent strain GCDE06; and Community 4) 1 mL of 10 5 cells mL 21 of a 5 mm filtrate from the non-axenic parent strain GCLV01. Two sets of experimental control cultures were also established as follows:  [25]. Survival and growth indicates bacterial contamination either from failed surface-sterilization, intracellular bacteria released at germination, or from other sources such as contaminated growth medium.
Petri dishes were incubated for 3 weeks at 21uC62uC, 80 mmol PAR m 22 sec 21 with a 18:6 h light:dark cycle to allow excystment of the resting cysts. Clonal isolates were established from the mixed offspring cultures by dilution in sterile GSe medium (to 1 cell mL 21 ) and aseptic distribution of 1 mL aliquots into wells of sterile 24-multiwell plates containing 2 mL sterile GSe medium. Wells were examined using a Leitz Labovert FS inverted microscope (2006 magnification), those containing a single cell/chain were labelled for later isolation, and the plate incubated for a further week. Clonal cultures were established from labelled wells by aseptic transfer into 50 mL Erlenmeyer flasks containing 25 mL GSe medium, and 2 weeks later to 50 mL flasks containing 40 mL GSe medium.
A total of 5 clonal isolates were established for each of the 5 bacterial treatments, together with 5 clonal cultures of each of the two parent strains (GCDE06 and GCLV01). The clonal cultures were maintained on three week aseptic transfer intervals in 50 mL Erlenmeyer flasks at 18uC, 80 mmol PAR m 22 sec 21 with an 18:6 h light:dark cycle (Phillips cool-white fluorescent). Transfers to fresh medium were made every 2 weeks.

Toxin production of cultures
Cultures for toxin content and production rate estimates were transferred to 175 mL Erlenmeyer flasks and maintained for 50 days under the same light and temperature conditions described earlier to establish the period of exponential growth phase of cultures prior to the toxin production experiment. At Day 0, 10 mL of late exponential phase cultures was inoculated into 150 mL of sterile GSe medium and the remaining inoculum culture volumes retained for triplicate cell counts (Leitz Labovert FS microscope, 2006 magnifications) and toxin analysis. Subsequent cultures were grown at 18uC and 80 mmol PAR m 22 sec 21 light with a 18:6 h light:dark cycle until Day 25 (late-exponential phase) and harvested for toxin analysis. Cell concentration was

HPLC analysis of PSTs
PSTs including C-toxins, gonyautoxins (GTX) and the hydroxybenzoate (GC) toxins were analysed by HPLC using the methods of Negri and Jones [28] and Negri et al. [20]. Briefly, toxins were separated using a Waters 600 HPLC, with postcolumn reactor (Pickering PCX 5100) using a 5 mm, 250 mm64.6 mm Alltima ODS column (Alltech, IL, USA) with a flow rate of 0.8 mL min 21 . Post-column oxidation was performed according to the method of Oshima et al. [4].
Derivative PST fluorescence was detected with a Linear LC305 spectrofluorometric detector (excitation at 330 nm and emission at 390 nm). The retention times and fluorescent intensity of the PSTs were compared with PST standards (NRC, Canada) and identity of PST compounds confirmed by sample spiking experiments and removing post-column oxidation and observing the disappearance of peaks. The quantification of toxins was achieved by comparing peak areas with those of authentic standards and combining this data with cell counts from the time of harvest then converting to total PST expressed as fmol cell 21 . As the proportion of a-epimers is often very small and some interconversion is expected, concentrations of epimer pairs were combined comparing toxin profiles.

Calculations and statistical analyses
Specific growth rates (m) of the cultures from Day 0 to Day 25 were calculated over the exponential growth phase (Day 0 to Day 25) using the equation:  where N1 and N2 = cell concentration at time 1 (t1) and time 2 (t2) respectively [29]. Specific toxin production rates (m tox ) and net toxin production rates between Day 0 and 25 were calculated using the equations and methods described by Anderson et al. [30]. Briefly, the STX content of cells (fmol cell 21 ) at Day 0 and at Day 25 was multiplied by N t (cell concentration at time t) to yield T t , the total toxin concentration (fmol STX mL 21 culture) at time t. Toxin concentrations (fmol STX mL 21 ) at Day 0 were calculated from the STX content (fmol cell 21 ) of the inocula and included a correction for dilution at inoculation. Values of T were then used to calculate specific toxin rate m tox over each time interval: The net toxin production rate R tox (fmol toxin cell 21 d 21 ) was determined using: where N' is the ln average of the cell concentration (below): and Dt ( = t 2 2t 1 ) the interval between Day 0 and Day 25. The ln average concentrations are used to account for exponential growth over the time period t 1 2t 2 [30]. The correlation between the specific toxin production rate and algal growth rate was done using linear regression. Significant differences in toxin production among bacterial culture treatments were compared using one-way ANOVA using the statistical software package R (Version 2.9.0), with treatment and time (Day 0 and 25) as factors, followed by Tukey's post hoc tests of significance. Differences in culture toxin profiles were examined using PERMANOVA+ [31], a multivariate analysis of variance with significance testing by random permutation, as implemented in the PRIMER-6 software package (www.primer-e.com). The toxin profile dataset (mol% STX, Day 25) was standardised prior to analysis to account for unequal variances among treatments. Canonical analysis of principal coordinates (CAP) [32] was performed using principal coordinates (PCO) calculated from a Euclidean distance matrix, and the first two canonical axes plotted. Over-parameterisation of the CAP was limited by restricting the number of PCO axes to that which maximised the leave-one-out allocation to groups [33]. All PERMANOVA and CAP tests used 9999 unrestricted random permutations of the raw data.

Culture growth dynamics
Both the positive and the sterility control cultures behaved as expected. The sterility control did not survive, indicating that growth media was not contaminated. The positive controls germinated (in the presence of G. catenatum representative bacterial community), survived and were used to establish five replicate cultures. The experimental cultures of G. catenatum exhibited average exponential growth rates (K') between Day 0 and 25 of 0.09560.005 to 0.1460.008 Div. day 21 with no significant variation among cultures grown with different bacterial communities ( Table 2) (F = 1.96, df = 6, p = 0.092).

Toxin content and production
There was no significant difference in the toxin content of the offspring cultures between Day 0 and 25, (F = 3.416, df = 4, p = 0.072) therefore the data from both time periods was pooled. Cultures established with mixed bacterial communities contained a significantly higher toxin content than cultures grown with single bacterial types (F = 5.89, df = 6, p = 0.007) (Fig. 1). The specific toxin production rate (m tox ; fmol day 21 ) varied from 0.07 to 0.13 fmol day 21 (F = 6.35, df = 34, p = 0.0003) (Fig. 1) and was strongly correlated with the exponential growth rate of the cultures (df = 34, p = ,0.001, R 2 = 0.82) (Fig. 2). Net toxin production rate (R tox ; amol cell 21 day 21 ) varied considerably among replicates and consequently no significant difference was detected among the different treatments (F = 2.44, df = 6, p = 0.064). However, the net toxin production rate of offspring cultures was almost 8-fold lower than the parent cultures (t = 2.83, df = 33, p = 0.008) (Fig. 3).
The CAP analysis of toxin profile data for all 35 cultures resolved three distinct toxin profile groupings. Parent cultures GCDE06 and GCLV01 were clearly separated along axis 2 (yaxis), primarily due to the different ratios of C1+2 versus C3+4 (Fig. 6). Cultures grown with Marinobacter sp. DG879 clustered separately along axis 1 (x-axis), and all other offspring cultures clustered with or near parent GCDE06, or between the GCDE06 and GCLV01 parent clusters (Fig. 6).

Discussion
Bacterial effects on toxicity of G. catenatum A number of previous studies have also shown that bacteria change or induce cellular toxicity of axenic and non-toxic algal strains [12], [18], [34], [35], [36] however our study is the first to show that altering the culture-associated bacterial community results in changes to both PST content and PST production rate. The toxin content of G. catenatum vegetative cells varies considerably [4], [21], [37] and is influenced by a range of culture-related and environmental factors such as isolation method, growth medium, nutrient type, concentration, salinity and temperature [21], [38], [39], [40]. All of these factors were controlled across our treatments, therefore our data shows that change in both cellular toxicity, net PST production and PST profile of the dinoflagellate is due to the modification of the bacterial community. Culture-based microbial diversity of the unibacterial cultures routinely recovered only the single bacterial type added in each case (Alcanivorax or Marinobacter) and we are confident that the cultivable communities were overwhelmingly dominated by the added bacterial strain.
Previous work shows that G. catenatum culture bacterial communities are typically composed of 15-24 distinct bacterial genotypes and a consistent community structure dominated by aand c-Proteobacteria and Bacteroidetes [11], [24]. While we do not have detailed microbial community data for the mixed community cultures in the present experiment, it is reasonable to assume similarly complex communities associated with parent and the mixed community offspring cultures. The observed changes in toxicity could then be related to changes in bacterial associate composition (i.e. the specific bacteria present), the reduction in community complexity, or both. Which factor is most important is difficult to determine from our experimental design and data, however it is interesting to note that the two uni-bacterial cultures showed the lowest cellular toxicity suggesting that reduced bacterial diversity may be partially responsible for reduced toxin content of G. catenatum. Axenic and antibiotic treated A. catenella and A. tamarense cultures with substantially reduced bacterial density and diversity also show reduced PST content [18], [41] indicating that this response is not confined to G. catenatum.
Earlier studies of G. catenatum have shown up to 40-fold reduction in the cellular toxicity of G. catenatum cultures established from wild resting cysts [21]. We did not observe the same scale of reduction in cyst germinated offspring, however, we did detect an 8-fold reduction in mean net toxin production rate (amol cell 21 day 21 ) compared to parent cultures. Studies of PST profiles of resting cysts indicate that PST production is reduced or stops during gamete and resting cyst formation [42], [43]. Sonication and washing of cysts during isolation may remove specific microbial associates (and their metabolites) required to fully induce dinoflagellate PST synthesis pathways after germination [21], [24]. Alternatively, reduced PST production may be associated with down-regulation of many secondary metabolite synthesis pathways during cyst formation and dormancy, or a form of reversible gene-silencing as described in some mycotoxinproducing fungi [44], [45].
The strong linear relationship between specific toxin production rate (m tox ) and specific growth rate (m), and the poor correlation between net toxin production rate (R tox ) and m found for G. catenatum in this study, is similar to that of Alexandrium fundyense [30]. Anderson et al. [30] suggested this pattern resulted from nutritional deficiencies in batch cultures, causing an uncoupling of toxin synthesis from cell division, leading to highly variable rates of toxin accumulation. We sought to minimise nutritional limitation by harvesting cultures while in logarithmic growth and using saturating light intensity for G. catenatum at 21uC (80-90 mmol PAR m 22 sec 21 ) [46], however, absolute growth rates observed in our experimental cultures are relatively low compared to other culture studies at similar temperatures and some of our cultures may have been light-limited leading to nutrient-limited growth in some cases. Alternatively, PST synthesis has been shown to occur only during the G1 phase of cell division [47], therefore the poor correlation may be due to differences in division synchrony among cultured treatments. Secondary metabolites (such as toxins) are also often subject to induction effects (up regulation) in response to unfavourable conditions [48], [49], further reducing the likelihood of a simple correlation with growth rate. Current knowledge of saxitoxin gene expression in dinoflagellates is limited, but recent studies suggest transcriptional or pre-translational regulations are major components. Taroncher-Oldenburg and Anderson [49] identified more than 20 transcriptionally regulated genes that were differentially expressed throughout the cell cycle of Alexandrium fundyense which were either up-or down regulated during toxigenesis. Similarly, Zhuang et al. [50] found that genes related to toxin production were expressed at different levels at different time points of the diel cycle.

Bacterial effects on toxin profile
The dominance in our cultures of sulfocarbamoyl toxins C1+2 and C3+4 and the more recently characterised hydroxy-benzoate toxins (GC1+2) is consistent with the range of toxins evident from earlier studies of Australian G. catenatum [4], [21]. The substantially different profile produced by cultures grown with Marinobacter sp. DG879 (Gc/Mar; lower C1+2 and C3+4; greater dcGTX2+3 and GC1+2) in comparison to the other cultures, could be due to conversion of C1 and C2 to their decarbamoyl derivatives dcGTX2 and dcGTX3. The GC-toxins, GC1 and GC2 are p-hydroxybenzoate analogues of the carbamate toxins GTX2 and GTX3 (respectively) but the biosynthetic pathway of GC-toxins is not yet fully understood [9], [10], [23].
Whether the observed PST congener transformations are carried out by G. catenatum only in the presence of Marinobacter sp. DG879 or by the bacterium itself is not clear. Bacterial PST production has been speculated or suggested by many studies [51], [52], [53] but there is currently no convincing structural evidence to confirm PST production by heterotrophic bacteria. Our previous studies have also shown that Marinobacter sp. DG879 does not produce PST-like toxins or activity [24] therefore the PST profile changes observed in GC/Mar cultures is not explained by production of particular PST congeners by Marinobacter. The recent discovery of dinoflagellate homologues of saxitoxin synthesis genes (e.g. SxtA1, SxtA4, SxtG) in Alexandrium, Pyrodinium bahamense and Gymnodinium catenatum [9], [10], [54] also appears to confirm that saxitoxin synthesis in our cultures is primarily or exclusively of dinoflagellate origin.
The changes may be mediated by bacterial enzymatic transformation of PSTs [14], [15], [16]. For example an uncharacterised carbamoylase activity has been demonstrated in bivalve-associated gut enzymes transform GC1, GC2 and GC3 to the more toxic dcGCX2, dcGTX3 and dcSTX [16]. This pathway could explain transformation within G. catenatum of C1+2 to dcGTX2+3, but does not explain the high proportions of GC1+2. On the other hand, proportional increase in GC1 and GC2 is balanced by a similar reduction in C1 and C2, suggesting separate biosynthetic pathways or more complex enzymatic transformations. To date all bacterial PST transformation activities in earlier studies have, however, been observed in supernatants of bacterial cultures or shellfish extracts. This indicates that activity is mediated by extracellular enzymes, presumably only acting on PST in the surrounding medium or cell boundary layer. We examined only intracellular PST in this work therefore any bacterial mediated PST transformation would need to be mediated by enzymes produced by intra-cellular bacteria or by surface bacteria releasing enzymes that are actively transported into the dinoflagellate cell. Alternatively particular bacterial associates may produce metabolites that have an indirect influence of Marinobacter sp. DG879 via their action on toxin synthesis pathways or physiology of the dinoflagellate cell.
Multivariate comparison of parent PST profiles with the 25 offspring from our experiments contradict previous parentoffspring studies of G. catenatum [5] and Alexandrium species [55], [56] which all report segregation of parental toxin profiles in a 2:2 Mendelian pattern. If this was the case for our study, the CAP ordination should show offspring clustering with or near the parent profiles in equal approximately numbers. Excluding the GC/Mar offspring that cluster separately on the axis 1 of the CAP, the majority of other offspring profiles cluster almost exclusively with parent GCDE06 (see Fig. 6) and indicate a uni-parental inheritance pattern in our experiment.
Resting cysts were not produced in self-crosses of either parent culture so it is highly unlikely that a majority of the offspring were products of self-crosses of GCDE06. A 2:2 segregation pattern also assumes the major determinants of toxin profile are inherited as single locus. Given the complexity of the core saxitoxin biosynthetic pathway (.20 genes), the additional predicted tailoring enzymes, and also a probable presence of multiple expressed copies [10], it would in retrospect be surprising to find that PST profiles were inherited as a single non-recombining unit in dinoflagellates.
Our controlled bacterial community manipulation and germination study demonstrates that interactions with bacteria result in distinct and reproducible changes in cellular toxicity and toxin production rate, and cause significant changes to the intracellular PST profile of G. catenatum. How the observed changes are mediated remains unclear, however the balance of evidence suggests bacterial influences on dinoflagellate cell physiology and growth are the most likely mechanism.
Author Contributions