*Article* **Asynchrony of** *Gambierdiscus* **spp. Abundance and Toxicity in the U.S. Virgin Islands: Implications for Monitoring and Management of Ciguatera**

**Justin D. Liefer 1,2, Mindy L. Richlen 3, Tyler B. Smith 4, Jennifer L. DeBose 5, Yixiao Xu 3,6, Donald M. Anderson <sup>3</sup> and Alison Robertson 5,7,\***


**Abstract:** Ciguatera poisoning (CP) poses a significant threat to ecosystem services and fishery resources in coastal communities. The CP-causative ciguatoxins (CTXs) are produced by benthic dinoflagellates including *Gambierdiscus* and *Fukuyoa* spp., and enter reef food webs via grazing on macroalgal substrates. In this study, we report on a 3-year monthly time series in St. Thomas, US Virgin Islands where *Gambierdiscus* spp. abundance and Caribbean-CTX toxicity in benthic samples were compared to key environmental factors, including temperature, salinity, nutrients, benthic cover, and physical data. We found that peak *Gambierdiscus* abundance occurred in summer while CTXspecific toxicity peaked in cooler months (February–May) when the mean water temperatures were approximately 26–28 ◦C. These trends were most evident at deeper offshore sites where macroalgal cover was highest year-round. Other environmental parameters were not correlated with the CTX variability observed over time. The asynchrony between *Gambierdiscus* spp. abundance and toxicity reflects potential differences in toxin cell quotas among *Gambierdiscus* species with concomitant variability in their abundances throughout the year. These results have significant implications for monitoring and management of benthic harmful algal blooms and highlights potential seasonal and highly-localized pulses in reef toxin loads that may be transferred to higher trophic levels.

**Keywords:** *Gambierdiscus*; ciguatera poisoning; *Dictyota*; ciguatoxin; Caribbean; dinoflagellate; benthic algae; algal toxin; harmful algal bloom

**Key Contribution:** This study demonstrates a seasonal asynchrony between *Gambierdiscus* abundance and C-CTX toxicity through monthly long-term monitoring, with the most significant trends observed at offshore field sites where depth and other factors may favor toxin production. This asynchronicity reflects potential differences in the toxin cell quota of individual *Gambierdiscus* species and their variations in relative abundance with the species assemblage through time. These data highlight the need for increased spatio-temporal monitoring focused on identifying seasonal and site-specific pulses in CTX production in order to estimate potential CP risk rather than depending on genus-level abundance data.

**Citation:** Liefer, J.D.; Richlen, M.L.; Smith, T.B.; DeBose, J.L.; Xu, Y.; Anderson, D.M.; Robertson, A. Asynchrony of *Gambierdiscus* spp. Abundance and Toxicity in the U.S. Virgin Islands: Implications for Monitoring and Management of Ciguatera. *Toxins* **2021**, *13*, 413. https://doi.org/10.3390/ toxins13060413

Received: 3 May 2021 Accepted: 7 June 2021 Published: 10 June 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

#### **1. Introduction**

Of all the human poisoning syndromes associated with harmful algal blooms (HABs), ciguatera poisoning (CP) has the most significant human health and economic impacts globally [1]. CP is caused by consumption of fish or shellfish, generally associated with coral reef systems, that are contaminated with a suite of lipid-soluble toxins known collectively as ciguatoxins. These toxins and their precursors are produced by certain species or strains of benthic dinoflagellates in the genera *Gambierdiscus* and *Fukuyoa* that live on algal substrates or other surfaces, such as dead corals and sand, in many coral reef communities [2–4]. Ciguatoxins (CTX) and related metabolites enter and accumulate in coral reef food webs through grazing by herbivorous fish and invertebrates, reaching their highest concentrations in carnivorous finfish [5–7], where they pose the greatest public health risk [8]. With increased globalization, CTX and *Gambierdiscus* have been reported from temperate locations including the northern Gulf of Mexico, New Zealand, Japan, and the Canary Islands; however, CP is endemic to many tropical and subtropical coral reef ecosystems globally and primarily affects coastal communities [9–12]. Global estimates of CP incidence vary widely, ranging from tens of thousands to as many as 500,000 poisonings per year [1,13]. Producing accurate estimates of the true incidence of CP is challenged by a high degree of underreporting and misdiagnoses [14], and consequently, CP remains an overlooked and under-appreciated problem. Prevention and management strategies for CP have been hindered by knowledge gaps regarding the environmental and physiological factors contributing to toxin dynamics, as well as the lack of commercially available toxin standards and affordable and practical methods for toxin detection.

CP differs from other HABs in that poisoning events are not associated with largescale planktonic blooms of a single causative species but are often an ongoing and chronic problem in endemic regions. Over the past two decades, renewed scientific interest and research has resulted in significant advances in our understanding of the biogeography and ecophysiology of *Gambierdiscus* and *Fukuyoa,* including a fuller characterization of species diversity and global distribution [2,15], intra- and inter-specific [16–19] growth characteristics [20,21], and habitat or substrate preferences (reviewed by [2]). These studies have provided a fuller understanding of the factors governing population and toxin dynamics, including the identification of highly toxic species that may dominate CTX production and flux into food webs. Key drivers of CTX prevalence and CP risk are thought to involve a combination of several environmental and ecological factors, including: (1) environmental conditions that promote growth, leading to high *Gambierdiscus* and *Fukuyoa* cell concentrations, (2) prevalence of CTX-producing species and strains, (3) environmental conditions that promote cellular toxin production, and (4) increased substrate availability that promotes the proliferation and increased areal abundance of benthic dinoflagellates in reef ecosystems. Additionally, patterns of toxicity are affected by bioconversion of the toxin precursors produced by *Gambierdiscus* to more potent compounds during toxin uptake, metabolism, and transfer.

As research on *Gambierdiscus* has expanded and progressed, so have efforts to characterize linkages between key environmental factors (e.g., seawater temperature), *Gambierdiscus* and *Fukuyoa* population and toxin dynamics, and subsequent CP risk and incidence. Early efforts by Tosteson [22] reported a relationship between warming seawater temperatures and barracuda toxicity, in which barracuda ciguatoxicity was correlated with both increases in seawater temperatures as well as with total cases of human ciguatera intoxications. In Tahiti, Chinain et al. [23] carried out weekly sampling of *Gambierdiscus* abundance and toxicity over a seven-year period (1993–1999), which identified seasonal trends in peak cell densities that occurred during the warmest months (October, November, December). They concluded that ciguatera outbreaks more likely reflected the presence of highly toxic strains rather than high overall biomass, as no correlation was found between sample toxicity and *Gambierdiscus* spp. abundance. These data were subsequently used by Chateau-Degat et al. [24] to construct a temporal model that related seawater temperatures to *Gambierdiscus* spp. growth, and subsequent onset of ciguatera cases. A key challenge

in these and other ongoing efforts to link spatiotemporal dynamics of *Gambierdiscus* spp. abundance and CTX production with CP risk is the high spatial heterogeneity observed for both *Gambierdiscus/Fukuyoa* populations and the ciguatoxicity of potential seafood vectors. Within small spatial scales (<3 km) on a single island, fish at one reef site may be safe to eat, while neighboring reefs can harbor "hot-spots" where ciguatoxic fish are prevalent [25–28]. This variation in CTX accumulation within higher trophic levels may reflect corresponding spatial heterogeneity in *Gambierdiscus* population structure, coupled with the large differences in CTX production documented among co-occurring *Gambierdiscus* and *Fukuyoa* species [9,16,17,29,30]. For example, there can be an over 1500-fold difference in toxin content among *Gambierdiscus* species, with *G. polynesiensis* identified as a toxic species from the Pacific [7], and *G. excentricus* and *G. silvae* as the most toxic species from the Caribbean [17,19,31,32].

In this study, we assessed *Gambierdiscus* abundance and CTX content within natural epiphyte assemblages from St. Thomas, USVI (Figure 1), an area hyperendemic for CP [33], to determine the seasonality, spatial variability, and environmental correlates of CTX production. Field activities were conducted monthly over 3 years at four coral reef sites located on the south side of the island, including two nearshore (Black Point, BP; Coculus Rock, CRK) and two offshore locations (Flat Cay, FC; Seahorse Shoal, SH), ranging in depth from ~6–22 m depth (see Figure 1). Populations of *Gambierdiscus* found at these sites are known to comprise five of the seven *Gambierdiscus* species documented in the Caribbean and one ribotype: *G. belizeanus*, *G. caribaeus*, *G. carolinianus*, *G. carpenteri*, *G. silvae,* and *G.* sp, *ribotype 2* (Richlen, M.L. unpublished data, and [34]). Clear, yet decoupled, seasonal patterns of *Gambierdiscus* abundance and CTX levels were observed, as well as marked differences in CTX levels among adjacent sites. Our findings indicate that variability in CTX production within *Gambierdiscus* populations over small spatial scales may be a key driver of CP risk. This work also highlights the importance of time-integrated monitoring of in situ CTX production, which provides a more direct determination of the sites and conditions that are the ultimate source of CP risk.

**Figure 1.** Map of long-term field sampling sites in St. Thomas, US Virgin Islands. Map created in ArcGIS Professional with

overlaid shapefiles of benthic cover with coral, hardbottom, and seagrass from [35] and macroalgae from the St. Thomas and St. John benthic habitat dataset [36], both from the National Oceanic and Atmospheric Administration, U.S. Dept. Commerce.

#### **2. Results**

#### *2.1. Variation in Environmental Conditions and Benthic Composition*

Mean daily temperature and salinity (not shown) were similar at all four sites and followed seasonal patterns. Benthic temperatures ranged from 25.5–29.9 ◦C with a mean of 27.8 ± 1.6 ◦C and displayed a seasonal pattern typical of the tropical northern hemisphere, with peak temperatures in summer (June–October) and minimum temperatures in winter (December–March). Salinity variations were small and ranged from 34.6–36.2 psu with a mean of 35.5 ± 0.5 psu, with maximum values in March–May and minimum values in September–November. Dissolved nutrients were low overall and varied over a small range. For instance, mean dissolved inorganic phosphorus (DIP) was 0.09 ± 0.04 μM, while mean dissolved inorganic nitrogen was 0.90 ± 0.8 μM. There were also no strong seasonal patterns in available data on climate variables (precipitation, wind speed and direction). Spatial and seasonal variation in environmental and physical parameters (i.e., wind speed and direction, precipitation, benthic temperature, and nutrients) were also examined with multivariate techniques and no clear ordination of parameters or clustering of sites was apparent (see Table S1). Analyses were limited to these parameters due to frequent gaps in other data (e.g., salinity and other CTD vertical profile measurements). Selected environmental parameters were compared using a principal components analysis (PCA) and no principal components explained more than 22.8% of overall variation (Table S1). Environmental variation among sites was examined with a cluster analysis, multidimensional scaling, and an ANOSIM test (Primer-E) for parameters that were measured directly at each site (benthic temperature and nutrients). All sites showed high multivariate similarity (ANOSIM Global R = 0.28) and no clear spatial patterns.

Benthic composition at all sites was mostly dominated by macroalgae, which ranged from 11.7–75.0% of benthic cover with a mean of 39.6 ± 12.5%, and was variable over time (Figure 2). Other major components were dead coral with turf algae, non-living substrate, and corals, with mean benthic cover ranging from 11.7–23.0%. Minor components included gorgonians, sponges, coralline algae, and cyanobacteria. As with the site environmental conditions, there were no clear spatiotemporal patterns and a high multivariate similarity (ANOSIM test, Primer-E) in overall benthic cover among sample years and study sites. Macroalgal cover was dominated by fleshy macroalgae which mainly consisted of *Dictyota* spp. (24.1 ± 10.0% of macroalgal cover; see monthly trends in Figure 2), followed by *Lobophora variegata* (7.7 ± 10.4%), and *Halimeda* spp. (0.4 ± 1.2%). Macroalgae composition was distinct among the sites (ANOSIM Global R = 0.568) (Supplementary Figure S1), primarily due to differences between the nearshore sites (CRK and BP) and the offshore Sites FC (R = 0.501–0.539) and SH (R = 0.824–0.935) (Table S1). A SIMPER analysis showed these dissimilarities were driven by the higher abundance of *L. variegata* at the offshore sites, which accounted for 41.7–60.7% of the dissimilarity between pairwise comparisons.

**Figure 2.** Time-series of benthic cover over the study period as determined by benthic habitat video surveys. Data highlights the temporal change in the percent (%) cover of combined macroalgae (black) and % *Dictyota* spp. cover (green) at each sampling site. Missing data points are time periods when benthic surveys were not conducted. Feb\* denotes that FC was surveyed on 23 February 2010.

#### *2.2. Gambierdiscus spp. Abundance*

*Gambierdiscus* spp. were detected in all 135 samples collected in this study. Abundances of *Gambierdiscus* were highly variable, both in terms of overall range (2.5–63.3 cells g *Dictyota*−1, mean 69.0 ± 63.3 cells g *Dictyota*−1) and periodicity (Figure 3). The only apparent seasonal pattern in abundance was the occurrence of an annual maximum in September–October of each study year at each site that coincided with the thermal maximum of sea surface temperature and doldrum-like conditions. The variation in *Gambierdiscus* spp. abundance was dissimilar among sites, with the exception of the annual abundance peaks in September–October of 2010 and 2012 (see Figure 3). The highest overall abundances at CRK, FC, and SH were observed in September–October 2010, and annual mean *Gambierdiscus* spp. abundance was also significantly higher (*p* < 0.001, ANOVA with Tukey's test post hoc) in 2010 (111.5 ± 80.8 cells g *Dictyota*−1) compared to 2011 and 2012 (49.4 ± 35.6 and 52.7 ± 51.3 cells g *Dictyota*−1, respectively). Mean abundances were generally higher at BP (84.7 ± 69.3 cells g *Dictyota*−1) and FC (80.4 ± 56.9 cells g *Dictyota*<sup>−</sup>1) near the western end of St. Thomas compared to the eastern sites of CRK and SH (56.6 ± 62.0 and 54.7 ± 61.6 cells g *Dictyota*−1, respectively), with overall abundance at BP being significantly higher than at CRK and SH (*p* < 0.05). There was no significant correlation between abundance and any of the environmental factors assessed, based on both direct parametric correlation tests and multivariate correlation analysis (BEST in Primer-E).

**Figure 3.** Asynchrony between cell toxin load (C-CTX eq cell−<sup>1</sup> \* cells g *Dictyota*−1) and mean (+s.d.) *Gambierdiscus* spp. abundance (cells g *Dictyota*<sup>−</sup>1), with 7-day averaged bottom temperatures from each site. Open circles represent "No Data" for either toxin load or *Gambierdiscus* abundance count.

#### *2.3. Detection of Ciguatoxins in Field Samples*

The specific sodium channel agonist activity detected by N2a assay was attributed to CTX congeners in all samples based on several lines of evidence: (1) the direction, shape, and slope of dose-response curves generated from field sample extracts were congruent with C-CTX-1 standards indicating similar activities and potencies by N2a assay (see Supplementary Figure S2); (2) *Gambierdiscus* spp. (a known source of CTXs) was present in all samples; (3) the extraction procedure used was not suitable for isolation of the polar alkaloid sodium channel blockers known to occur in marine systems (e.g., saxitoxin, tetrodotoxin); (4) non-specific activity potentially generated by other toxin classes (with alternate modes of action) were excluded from our analyses (as described in the Methods); and, (5) dinoflagellate sources and toxins of other site 5 sodium channel agonists, e.g., *Karenia*-produced brevetoxins, have not been reported in algae, fish, or shellfish from the study region. Considering the biotransformation of CTXs characterized in other regions, the CTX congeners detected in Caribbean *Gambierdiscus* spp. are likely to be uncharacterized precursors of C-CTX-1 or C-CTX-2, the most abundant congeners found in higher trophic level Caribbean fish [37,38]. The Caribbean CTX standard used in the N2a bioassays was C-CTX-1, the only quantified reference material that was available at the time of this study, hence all detected CTX levels are expressed as C-CTX-1 equivalents (C-CTX eq.).

The identity, structure, and toxicity of Caribbean CTX congeners present in *Gambierdiscus* spp. are poorly understood and no reference materials are presently available. This lack of knowledge complicates the use of clean-up methods, such as solid-phase extraction (SPE), for sample extracts as they may remove the target analytes. As detailed in the methods, four samples, representing a range of determined CTX activities, were purified using silica (Si) SPE column (Agilent) to assess the effect of sample clean up on measured composite toxicity. In all four samples, Si SPE purification caused a reduction in assay response in both the ouabain-veratridine-treated cells (i.e., reduction in CTXs) and PBS control cells (i.e., a reduction in cytotoxic matrix compounds). This indicates that Si SPE clean-up may have removed cytotoxic matrix compounds that affect negative control N2a cells, but also removed some of the target analyte with no improvement in quantification and hence was not used for sample quantitation. Dilution of samples (reducing both interfering matrix and analyte) that did not undergo clean-up or purification resulted in a 7–16% variation in quantitation. Considering the lack of precision in cell-based and other bioassays, as compared to instrumental methods (e.g., LC-MS/MS), this was considered an acceptable degree of variation.

CTX activity was quantifiable in 24.6% of benthic algal samples without purification while also meeting quality assessment controls of the N2a assay (summarized in Table 1). In the vast majority of remaining samples, CTX activity was below the limit of quantitation. The lowest CTX concentration quantified in our samples was 0.33 ± 0.06 ng C-CTX-1 eq. mL<sup>−</sup>1, which was above the determined limit of quantitation for CTX in unpurified extracts of environmental algal samples using the N2a assay (see Section 4.4.4).

**Table 1.** The distribution of positive and quantifiable samples as well as the mean and range of cell toxin quota and toxin load across all sites.


#### *2.4. Spatial and Temporal Variability of In Situ Ciguatoxin Levels*

Among the 135 sampling events, 125 samples were available for assessing CTX production by *Gambierdiscus*. Among these, 46 (37.7%) were positive for CTX activity (Table 1), meaning that there was ≤50% survival in ouabain-veratridine-treated (i.e., sensitized to sodium channel agonist) N2a cells and ≥95% survival in untreated N2a cells at the same dosage. CTX levels were high enough to be quantified in 30 (24.6%) samples (Table 1). Samples deemed positive for CTX (specific activity for a sodium channel agonist), but not meeting quantitation criteria, were considered "trace" detections. Three samples were also considered trace measurements due to detection of CTX in the toxin samples and a

corresponding detection of *Gambierdiscus* spp. in the abundance samples at the same sampling points, but a measurement of *Gambierdiscus* cell abundance in the samples collected to assess toxicity was not available. Cell toxin quotas for *Gambierdiscus* spp. were calculated by normalizing the measured CTX concentration to the *Gambierdiscus* cell abundance measured within a given toxin sample. The toxin load for each sample was calculated as the product of the cell toxin quota and the *Gambierdiscus* spp. abundance measured on *Dictyota* hosts (i.e., cells per g *Dictyota*) during the same sample collection. This toxin load represents the amount of toxin present per mass of macroalgal substrate (units of pg C-CTX-1 eq. g *Dictyota*−1) and is used as a proxy for the amount of toxin available for trophic transfer during each sampling event.

The toxin quota of *Gambierdiscus* ranged from 0–12.6 pg C-CTX-1 eq. cell−<sup>1</sup> (Table 1; Figure 4) and toxin load ranged from 0–453.8 pg C-CTX-1 eq. g *Dictyota*−<sup>1</sup> (Table 1; Figure 3). Unlike *Gambierdiscus* spp. abundance, both toxin quota and toxin load appeared to have a distinct annual seasonality. The majority of positive samples (80.0%) and quantifiable toxin samples (76.6%) were collected during February–June of each sample year (Figures 3 and 4). The six quantifiable toxin samples observed outside of this February–June season were collected from Sites FC and SH in July–August 2011 and January–February 2012, months adjacent to the greatest periods of toxin occurrence (February–June 2011 and 2012). Only four trace toxin detections were observed in September–November of all sample years. There were no significant differences in mean toxin quota or toxin load between sample years (*p* = 0.057–0.061), although the largest proportion of quantifiable toxin samples occurred in 2012 (50.0%) followed by 2011 (33.3%) and 2010 (16.7%).

The majority of positive (67.2%) and quantifiable samples (76.7%) originated from the offshore sites FC and SH, with 46.7% of quantifiable samples originating from site FC alone and 30% from site SH (Table 1). The highest toxin quotas were detected at nearshore sites CRK (12.6 pg C-CTX-1 eq. cell−1) and BP (8.4 pg C-CTX-1 eq. cell−1), though these values were far higher than the other, infrequent toxin detections at these sites (Figure 4). A Welch's ANOVA (a one-way ANOVA that assumes unequal variance) showed a significant difference in toxin quota (*p* < 0.005) and toxin load (*p* < 0.01) among sites. A post hoc Games–Howell test showed that both toxin quota (*p* < 0.01) and toxin load (*p* < 0.05–0.01) were significantly higher at FC compared to CRK and BP (Figure 5). Considering both spatial and temporal variability, 50.0% of positive samples and 63.3% of quantified samples occurred at offshore sites and during the high toxicity period of February–June.

There was no strong or significant correlation between toxin quota or toxin load and any of the environmental factors assessed, based on both direct parametric correlation tests and multivariate correlation analysis (BEST in Primer-E). This is not surprising considering that CTX was not quantifiable in ~75% of samples and the lack of ordination or apparent structure in the environmental data. Although no monitored environmental variables were directly correlated with toxin parameters, the vast majority of positive (69.6%) and quantifiable (90.0%) samples were collected when temperatures were below the mean of the study period (27.8 ± 1.6 ◦C). Additionally, greater majorities of positive (91.3%) and quantifiable (96.7%) samples were collected when salinity was above the mean of the study period (35.5 ± 0.5 ◦C).

The toxin quota of the *Gambierdiscus* present, rather than *Gambierdiscus* abundance, appeared to determine toxin load throughout the study. There was no correlation between *Gambierdiscus* abundance (Spearman's R = −0.090, *p* = 0.51) and toxin load overall (Spearman's R = −0.054, *p* = 0.68) Additionally, there were only four trace detections during the months of peak abundance in each year (September–November), while some of the highest toxin quotas and loads occurred when abundances were relatively low (Figure 3).

**Figure 4.** Time series of toxin cell quota (pg C-CTX-1 eq. cell<sup>−</sup>1) determined from benthic microalgal (20–200 μM fraction) field samples collected monthly from nearshore (Black Point; Coculus Rock) and offshore sites (Flat Cay; Seahorse Shoal) of St. Thomas, Virgin Islands. Black Point and Flat Cay (black bars) are western sites, whereas Coculus Rock and Seahorse Shoal (blue bars) are eastern sites. Colored circles represent "No Data" collected from the corresponding sites, with all other zeros indicating true non-detections of toxicity.

**Figure 5.** Truncated violin plot of log-transformed toxin load. Medians (dashed) and quartiles (dotted) shown. Trace detections were designated at toxin load 0.25 pg C-CTX-1 eq. g *Dictyota*−<sup>1</sup> (log-transformed to <sup>−</sup>0.6). Letters on plot indicate significant differences in toxin load between sites.

#### **3. Discussion**

Laboratory studies of isolated toxic microalgae are essential for confirming their toxicity and mechanisms of toxin production, but these measurements often differ from the levels of toxin production observed in natural systems. Thus, robust assessments of in situ toxin production by harmful microalgae are critical for understanding their true toxin dynamics and potential threat to ecosystem function or public health. This work is the most comprehensive and quantitative assessment to date of the in situ toxicity of *Gambierdiscus* in the Caribbean, the ultimate cause of more cases of human illnesses than any other harmful alga [1]. We also express this in situ toxicity in proportion to the mass of an abundant benthic substrate (the macroalga *Dictyota* spp.) consumed by potential vectors of CTX, providing a quantitative link between toxin production and trophic transfer. As these values were determined in a location where CP is endemic with monthly sampling over 3 years, our findings provide valuable constraints for efforts to model in situ CTX levels and the trophic transfer of CTX in the Greater Caribbean region.

#### *3.1. Relative Toxicity of In Situ Gambierdiscus*

The cell toxin quotas determined in this study, ranging from 0–12.62 pg C-CTX-1 eq. cell−<sup>1</sup> with a mean of 0.5 ± 1.63 pg C-CTX-1 eq. cell−1, are comparable to the limited number of other available in situ values (see Table 2). Values in Table 2 that were determined with the mouse bioassay, originally compiled as mouse units by Litaker et al. [15], were converted to composite CTX toxin quotas by assuming that one mouse unit is equivalent to 18 ng of CTX3C for Pacific samples and 72 ng of C-CTX-1 for Caribbean samples [39,40]. Though comparable to other in situ values, the toxin quotas we observed are also considerably higher than those of most cultured *Gambierdiscus* strains of Caribbean, eastern Atlantic, or Pacific origin measured using similar applications of the N2a assay as used in this study [17,19,41]. Most of these N2a toxin quotas were determined as CTX3C equivalents [17,19], a common Pacific congener of CTX that has been reported to be 2-fold more toxic than the Caribbean congener we used as a reference standard, C-CTX-1 [42]. This difference in standard toxicity could cause a lower CTX content to be determined in Caribbean strains measured with CTX3C as a standard rather than C-CTX-1. Even when taking this difference into account, the only Caribbean strains with toxin quota values comparable to the in situ values measured in this study are those from the species *G. silvae* (2.1–4.8 pg C-CTX-1 eq. cell−<sup>1</sup> [31]) and *G. excentricus* (0.47 pg CTX3C eq. cell−<sup>1</sup> and 1.43 pg CTX3C eq. cell<sup>−</sup>1, respectively [17,19]).


**Table 2.** Published values of in situ *Gambierdiscus* toxin quota. Toxin measurements determined by mouse bioassay (MBA) were originally compiled by Litaker et al. [15] and were standardized prior to conversion to toxin quotas, assuming one mouse unit = 18 ng of CTX3C and 72 ng of C-CTX-1 for Pacific and Caribbean samples, respectively [39,40].


**Table 2.** *Cont.*

<sup>1</sup> Mouse Bioassay (MBA); <sup>2</sup> Radioligand Receptor Binding Assay (RBA), <sup>3</sup> In vitro mouse neuroblastoma MTT based assay (N2a); <sup>4</sup> sample number (*N*). ˆ Caribbean Region.

#### *3.2. Cellular CTX Quota and Not Gambierdiscus Abundance Determines CTX Production*

One of the most notable findings of this study is the greater influence of toxin quota rather than *Gambierdiscus* abundance on CTX load overall and within any given site. The vast majority of CTX detections occurred from February to June of each study year when *Gambierdiscus* abundance was relatively low, while CTX was generally not detected during periods with the highest *Gambierdiscus* abundances. The only other study to our knowledge that has monitored site-specific *Gambierdiscus* abundance and in situ toxicity over time [23] found a similar asynchrony between abundance and toxicity, with the highest in situ toxin measurements observed at relatively low *Gambierdiscus* abundance, and *Gambierdiscus* abundance being a poor predictor of in situ toxicity. This decoupling or asynchrony of *Gambierdiscus* abundance and CTX production is consistent with wide variation in toxicity that has been observed among *Gambierdiscus* species in all regions where the genus is endemic [7,17,19,30,32]. The far greater variation in toxicity among species compared to the variation within a species [19,41,52–54] indicates that CTX source levels are mostly determined by species composition rather than *Gambierdiscus* abundance at the genus level [17]. Additionally, the species that appear to be most common and widespread in the Caribbean, such as *G. caribaeus*, *G. carolinianus, G. belizeanus*, and *G. carpenteri*, have low toxicities [17,19], thus periods when these species are abundant may not be expected to result in high CTX levels. In contrast, Caribbean species such as *G. silvae* have been shown through our prior efforts to have high toxin quotas [31], while others have reported similar trends for *G. excentricus* [17,19]. In both cases, the reported cell toxin quotas reported in these high toxin-producing strains of the respective species, could have the capacity to produce the in situ CTX loads observed in this study, even at low abundances. The discovery of "super-producing" strains of *G. polynesiensis* in Pacific waters has generated the hypothesis that a small relative abundance of highly toxic *Gambierdiscus* species may dominate CTX production that leads to CP outbreaks [7,40,48]. The high CTX loads observed at relatively low in situ *Gambierdiscus* abundances, the high toxicity of some *Gambierdiscus* species, and the low toxicity of the most common *Gambierdiscus* species in our study region all support the hypothesis that highly toxic, low-abundance species of *Gambierdiscus* dominate CTX production in the Caribbean.

#### *3.3. Nearshore vs. Offshore Sites*

*Gambierdiscus* CTX production also displayed spatial patterns that were generally consistent across three years of monitoring. *Gambierdiscus* was present in all samples collected over the study period, yet the majority of samples containing quantifiable CTX were collected at offshore sites (FC and SH) during February–June. Though distinct in their contribution to regional CTX levels, sites FC and SH were similar to nearshore sites with respect to measured physical and chemical features. Despite these apparent similarities, depth may be a factor that distinguishes the high CTX offshore sites (FC and SH), where sampling depths were greater (~18 m bottom depth) than at the nearshore sites CRK and

BP (~9 m bottom depth). This difference in depth may generate distinct light and water motion conditions for epiphytic *Gambierdiscus* populations between offshore and nearshore sites, the measurement of which were beyond the scope of this work. Measurements of light availability or optical qualities of the overlying water column were not available for the samples used in this study and physical conditions at all study sites were inferred from current and wave activity at one nearby buoy-monitoring location. However, water motion has been shown to be low, with little effect on *Gambierdiscus* populations in this study area [55], and is considerably lower than at locations where depth and water motion seem to affect *Gambierdiscus* abundance within epiphytic communities [56]. There is limited and conflicting evidence as to the physiological effect of light on toxin production within *Gambierdiscus* species or strains [44,57], but there are distinct growth-irradiance responses among co-occurring Caribbean *Gambierdiscus* species [20,21]. More favorable light conditions for constitutively more toxic species at deeper offshore sites could support the observed spatial patterns in CTX production. Future studies could test this supposition by determining if the most toxic species found at the offshore locations in this study or similar locations have lower optimal or maximum growth irradiances or are better adapted to the likely lower variability in irradiance of deeper benthic habitats.

Nearshore and offshore sites also varied in terms of the relative abundance and composition of macroalgae substrate. At the nearshore sites (BP and CRK), *Dictyota* spp. made up the majority of the macroalgal substrate for *Gambierdiscus* attachment. However, at the offshore sites (FC and SH), there was generally a higher percent-cover of fleshy macroalgae, as well as a greater proportion of species other than *Dictyota* present (e.g., *Lobophora variegata*). Although, beyond this study, to differentiate, there are multiple aspects where macroalgal abundance and composition might impact *Gambierdiscus* abundance and ciguatoxin transferability. Specifically, *Gambierdiscus*–macroalgae host interactions, which can be species-specific, might depend on shading potential, chemical cues, and hostpalatability to higher trophic levels [58–60]. Since *Gambierdiscus* species determination has further developed since this study, future studies could determine if there are particular in situ host associations or macroalgal abundance that favor particularly toxic species/strains and if these associations are linked to the environmental conditions and benthic community compositions that vary between these nearshore and offshore sites.

#### *3.4. Seasonality of CTX Production*

Seasonal patterns of *Gambierdiscus* CTX quota and CTX loads were observed, with the vast majority (80.0%) of CTX detections occurring in February–June, which provides key insights into the environmental and ecological factors controlling CTX exposure risk. A similar seasonal pattern for in situ toxicity was also observed by Chinain et al. [23] in Tahiti, with the majority of high in situ CTX levels observed at temperatures below annual means and low or no in situ toxicity observed in the warmest months. If CTX loads are indeed determined by highly toxic species occurring at relatively low abundances, then the temporal patterns observed in this study indicate that these species show seasonality in their occurrence. Of the wide set of monitored environmental conditions that may affect *Gambierdiscus* populations, temperature and salinity showed the strongest seasonal patterns. However, neither of these factors were directly correlated with CTX quotas or loads. Salinity varied over a relatively small range (34.6–36.2 psu), which is unlikely to have an effect among species or a physiological impact within a species [20]. Temperature showed a considerably larger seasonal variation (25.5–29.9 ◦C) that spans the known range of temperature optima for *Gambierdiscus* species [20,21]. Despite the lack of a direct correlation between temperature and CTX levels, it is striking that the vast majority of positive (69.6%) and quantifiable (90.0%) samples were detected when benthic temperatures were below the mean temperature of the study area (27.8 ± 1.6 ◦C). This mean temperature is also well above the growth optimum for *G. silvae*, the most toxic Caribbean species examined by Xu et al. [20], which had the lowest upper temperature limit for growth (29.8 ◦C) among eight *Gambierdiscus* species. Additionally, the strain of *G. excentricus* that

has produced the highest CTX quotas for an Atlantic species to date [19] was isolated from waters off the Canary Islands with relatively cool temperatures for *Gambierdiscus* (18–24 ◦C). The strain of *G. excentricus* that has produced the highest CTX quotas for a Caribbean species [17] was isolated from Pulley Ridge, located ~150 km offshore of Florida at a depth of 60–80 m where temperatures would be considerably lower than the mean temperature observed in the present study. The adaptation of the most toxic Caribbean strains to cooler temperatures is consistent with CTX detections being restricted to belowaverage temperatures in this study and may provide a key environmental constraint on CTX exposure risk.

#### *3.5. Implications for Assessing CTX Exposure Risk*

Our observation that CTX source levels in St. Thomas are determined by the toxin quota of *Gambierdiscus* cells rather than their abundance at the genus level has implications for efforts to predict CP risk in regions where this illness is endemic. Many proposed management efforts or models of potential CP risk are based on monitoring or predicting overall *Gambierdiscus* abundance [24,61–63]. Our findings and the apparent importance of species composition in determining the CTX production of a *Gambierdiscus* population [7,17] indicate that monitoring *Gambierdiscus* abundance alone would not help determine when and where trophic systems are likely to encounter and biomagnify CTX. Determining the most toxic *Gambierdiscus* species in an endemic CP location and using new molecular identification tools [34,64,65] to determine their spatiotemporal distribution may be more conducive to estimating CP exposure risk.

Even if the occurrence of CTX in the first trophic level could be accurately predicted in systems that yield ciguatoxic fish, predicting the spatial and temporal links between algal CTX production, bioaccumulation of CTX in higher trophic levels, and potential human exposure remain challenging. Our findings provide the basis for linking these phenomena by demonstrating that CTX production is restricted both spatially, within a relatively small study area (offshore sites in St. Thomas), and seasonally (~February–June). Determining the locations most likely to produce CTX allows studies of trophic dynamics of CTX (e.g., [26]) or of the site fidelity of key CTX vectors like large, mobile fish species to be related to a limited spatial source of CTX. By establishing a time-frame when CTX vectors are most likely to consume highly toxic *Gambierdiscus,* the lag between CTX production and potential human exposure can be better assessed. The restriction of CTX production to below-average temperatures and the possible importance of cool-adapted highly toxic species [8] also suggests lower temperature limits and a broader potential geographic range for CP risk in shallow marine habitats than previously estimated [63]. These implications (i.e., potential range expansion of cool-adapted toxigenic *Gambierdiscus* species) have also been suggested by others in the field [8,66,67], highlighting the importance of further evaluation so that monitoring and predictive models meet the needs of future risk assessment.

#### **4. Materials and Methods**

#### *4.1. Site Descriptions*

Samples were collected at four sites around St. Thomas (Figure 1) between late February/early March 2010 and December 2012. All St. Thomas sites are located south of the island on a nearshore to offshore gradient. Coculus Rock (CRK; 18.31257 N, 64.86058 W) is located near an emergent rock reef and is composed of diverse scattered stony corals on bedrock (6–7 m depth). Black Point (BP; 18.3445 N, 64.98595 W) is a nearshore fringing coral reef (7–16 m depth). Flat Cay (FC; 18.31822 N, 64.99104 W) is a fringing coral reef on the leeward side of a small uninhabited island (11–16 m depth). Seahorse Shoal (SH; 18.29467 N, 64.8675 W) is a deep patch reef 2 km offshore of St. Thomas (19–22 m depth). The latter three sites are star coral (*Orbicella* spp.) reefs with diverse coral and sponge communities. Further site descriptions can be found in [68]. During the sampling period for this study, these sites were impacted by a moderate thermal stress and coral bleaching event in August 2010 (widespread colony paling and bleaching, but limited mortality), which was

truncated by the passage of Hurricane Earl on 3 August [69,70]. This storm passed about 105 km NE of St. Thomas, causing wind gusts of up to 120 km hr−<sup>1</sup> and rainfall of 7.6 cm at the St. Thomas airport (see https://www.nhc.noaa.gov/data/tcr/AL072010\_Earl.pdf, accessed on 22 April 2021).

#### *4.2. Environmental Sampling*

#### 4.2.1. Oceanographic Measurements

Salinity measurements were obtained at each site from vertical profiles taken with a shallow-water Seabird SBE 25 recording at 8 Hz (Sea-Bird Electronics, Bellevue, WA, USA). Sensors were factory-calibrated within one year of deployment. Casts were made at anchor or on drift within 100 m (horizontal) of the research site. Casts were made within 1 m of the seafloor. Resulting data files were trimmed to the bottom meter of the downcast and averaged over this meter for use in analysis as this reflected the closest point to the reef organisms. Additional physical data was retrieved from the Caribbean Ocean Observing System St. John Oceanographic Buoy (VI 104; 18◦15.09 N, 64◦46.02 W; https://www.caricoos.org/station/st-john/us, accessed on 22 March 2021).

#### 4.2.2. Benthic Temperatures

Benthic temperatures were taken at each site by a shaded Hobo Water Temperature Pro V2, Onset Computer Corp., Bourne, MA, USA) affixed to a steel rod within 20 cm of the reef surface following prior methods [69,71]. Probes were calibration-checked prior to deployment in an ice bath and took readings every 15 min. Data were averaged over each day to determine 1-day mean, and further averaged over 7, 14, 21, and 30 days for the respective means for use in analyses.

#### 4.2.3. Precipitation

Precipitation and wind data for the region was recorded at the St. Thomas Cyril E. King Airport by a US National Weather Service station (TIST) and data was accessed at the National Climate Data Center https://www.ncdc.noaa.gov/cdo-web/datasets/GHCND/ stations/GHCND:VQW00011640/detail, accessed on 22 March 2021). The mean daily precipitation for the 14 days prior to a sampling event was calculated from daily summaries.

#### 4.2.4. Nutrient Analyses

Water samples for nutrient analyses were collected in whirlpak bags and stored on ice until return to the University of the Virgin Islands (UVI; within 5 h). Once at the laboratory, samples were transferred to acid-washed, sample-rinsed polypropylene bottles and frozen at −20 ◦C. Samples were shipped frozen to Woods Hole Oceanographic Institute (WHOI) and analyzed for inorganic nitrate plus nitrite (hereafter termed "nitrate"), ammonium, silicate, and phosphate using a Lachat Instruments QuickChem 800 fourchannel continuous flow injection system. This method is USEPA approved for nutrient analysis ranging from groundwater to the open ocean.

#### 4.2.5. Benthic Community Composition

Benthic cover at each study site was estimated using digital video along six randomly sited permanent transects as described in [71]. Each transect was 10 m in length and marked with steel rods, with transects spaced at least 3 m apart. Digital video was recorded perpendicular to the substrate and resultant images were cut into non-overlapping images, typically 15 per transect. Fifteen random points were placed on the image using Coral Point Count software [72] and characterized to the lowest identifiable taxonomic or abiotic level by a trained expert. Cover of each category (i.e., Coral, Gorgonians, Sponges, Zoanthids, Macroalgae, Coralline Algae, Dead Coral with Turf Algae, Non-Living Substrate, Other Living) was calculated for each transect by dividing the number of occurrences by the total number of points surveyed. Macroalgal cover was partitioned into % fleshy macroalgae, % *Dictyota* spp., % *Lobophora variegata*, % *Halimeda* spp., and % other.

#### *4.3. Biological Sampling*

#### 4.3.1. Collection of *Gambierdiscus* Epiphytes

*Gambierdiscus* were collected as epiphytes on macroalgae to determine their abundance and toxin content and scientific collection permits for this project were approved by the Virgin Islands Department of Fish and Wildlife, Marine Resources Division.

*Dictyota* spp. were the most widely distributed algae at the sampling sites (and frequently was the only macroalgal taxa present), so only *Dictyota* spp. were sampled for this study. Eight replicate samples of *Dictyota* spp. (four for *Gambierdiscus* abundance and four for toxin measurements) were collected by SCUBA divers from each study site in each month of the study period, with these exceptions: February 2010—only Flat Cay was sampled, March 2010—only Coculus Rock and Seahorse Shoal were sampled, and September 2012—no sites were sampled. Multiple thalli of *Dictyota* spp. were collected by carefully cropping and transferring to a Ziploc bag, which was then sealed underwater. Samples were stored in a cooler until processing within the same day. For sample processing, macroalgae were vigorously shaken for at least one minute to loosen the dinoflagellates, which were then sieved sequentially using 200 μm and 20 μm nitex sieves. *Dictyota* spp. retained in the 200 μm filter were removed, blotted dry with a paper towel, and weighed. With samples for *Gambierdiscus* abundance, the fraction of material retained on the 20 μm sieve was rinsed into a 15 mL conical tube, brought up to 10 mL with filtered seawater, and preserved with 0.5 mL formalin. For toxin samples, material retained on the 20 μm sieve was pooled from all four samples and rinsed into a shallow tray. The tray was maintained under low light (cool white fluorescent) and larger particulate material was allowed to settle to the bottom of the tray while living, detached *Gambierdiscus* cells (or other motile epiphytes) would remain in the overlying water due to active swimming and phototaxis. This overlying water was gently siphoned off and sieved again with a 20 μm sieve. Material collected with this final sieving was rinsed into 50 mL polypropylene centrifuge tubes (total capacity 60 mL) with filtered seawater and samples centrifuged at low speed (<1000× *g*) for 5 min to pellet cells. A small volume of overlying seawater was discarded to reach a final volume of 50 mL for each sample. This sample was then inverted several times to mix and a 1 mL aliquot was collected and preserved as described above for *Gambierdiscus* abundance measurements to determine the *Gambierdiscus* cell density within the toxin sample. After low-speed centrifugation of the remaining cell suspension and subsequent removal of supernatant, the cell pellet was stored at −20 ◦C or on dry ice (during shipping to Dauphin Island) prior to toxin analyses.

#### 4.3.2. *Gambierdiscus* Cell Enumeration

Preserved samples were gently shaken and 0.5–1.0 mL was loaded in a Sedgewick Rafter slide. *Gambierdiscus* cells were identified to genus based on cell size and shape using photomicrographs and line drawings, e.g., [73]. *Gambierdiscus* abundance was enumerated using a Zeiss Axioskop microscope at 100× magnification. Sample cell densities were determined by multiplying the summed cell counts by a subsample proportion factor, and then dividing this value by the *Dictyota* wet weight to express *Gambierdiscus* cell concentrations as cells g ww−1. At the time this study was conducted, methods for discriminating *Gambierdiscus* species had not yet been developed, so cell counts are given for total *Gambierdiscus* spp.

#### *4.4. Toxin Extraction and Analyses*

#### 4.4.1. Cell Pellet Extraction

Pellet material representing the 20–200 μm fraction of epiphytic material collected from *Dictyota* spp. (approx. 5 g per tube) were initially extracted in 10 mL of 100% methanol (MeOH) with 2 min. vortex mixing and 2 min. probe sonication on ice (5 s pulses, 20% power). The extract was centrifuged (3000× *g* for 5 min. at 20 ◦C) and the supernatant was collected. The sample pellet was then extracted two more times as before, but without additional probe sonication. Supernatants were pooled (30 mL total), diluted with water to

60% aqueous MeOH, and then partitioned three times with 25 mL dichloromethane (DCM). The recovered DCM fractions were pooled and dried by rotary evaporation at 30 ◦C. The sample residue was then quantitatively transferred from the evaporation flask with washes of MeOH and DCM, added to a 13 × 100 mm glass vial, and dried under high-purity nitrogen gas at 30 ◦C. Sample CTX residue was then redissolved in 100% MeOH (5 mL) with 2 min. vortex mixing and 2 min. bath sonication and stored at −20 ◦C until analysis. All extractions were performed with HPLC-MS grade solvents (Sigma) and ultra-pure (18 MΩ) water.

#### 4.4.2. Quantitation of CTX by In Vitro N2a Cytotoxicity Assay

The ciguatoxin content of each sample was measured as composite toxicity using an ouabain-veratridine (O/V) dependent in vitro neuroblastoma cytotoxicity assay (N2a assay) [74]. These assays utilized mouse Neuro-2a cells (ATCC, CL131; N2a), which were propagated and maintained under continuous growth as previously described [27,75]. Cells were harvested at 85–90% confluency and seeded to sterile 96-well polystyrene plates at a density of 4 × 104 cells well−1. The N2a assay measures sample toxicity as a loss in viability of N2a cells that have been sensitized with O/V, making these cell responses highly specific to sodium channel toxins (e.g., CTX) and thus adds a line of evidence for CTX (or a composite of CTXs) being present in samples when loss in viability is observed. Within each assay, the response of untreated N2a cells (serving as negative control) is assessed at the same sample doses provided to O/V-treated cells to determine if the sample contains other toxic substances that are not sodium channel toxins and could affect viability of O/V-treated cells.

For quantitative assays of CTX, triplicate dose-response curves were determined for both O/V-treated and untreated N2a cells exposed to eight concentrations of sample CTX residues redissolved by high-speed vortex in 100 μL of 5%-FBS-RPMI media spanning a 128-fold range. After 24 h of exposure to sample extracts, N2a cell viability is assessed as the reduction of 3-(4,5- dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) by metabolically active cells to a purple formazan product that is measured by absorbance at 570 nm. The concentration of sample extract at which 50% of N2a cells lost viability (IC50) was compared to the IC50 of a purified C-CTX-1 standard (50 pg starting dose) that was measured in a concurrent N2a assay seeded with the same batch of N2a cells seeded to sample assays. These analyses were possible due to an aliquot of purified C-CTX-1 stock that was purified from toxic *Sphyraena barracuda* harvested from the Virgin Islands and is the same FDA stock reported in several prior studies [e.g., 27, 75, and others]. Impurities were assessed by LC-MS/MS analyses prior to use and original stocks quantified via NMR and gravimetric analysis (data not available). The toxin content of samples is expressed as the mass of C-CTX-1 equivalents in *Gambierdiscus* cells (pg C-CTX-1 eq. cell<sup>−</sup>1) based on the toxin sample cell counts described above. Samples in which the amount of extract required to cause a 50% loss in viability of O/V-treated N2a cells, and also caused significant loss in viability in untreated N2a cells, were considered below the limit of detection. Samples like these, in which there is clear CTX-specific activity, but quantitation criteria are not met, were considered positive detections of CTX, but their CTX content is described as "trace".

#### 4.4.3. Sample Screening and Dose Determination

To determine which samples contained sufficient CTX-specific toxicity to be quantitated using the procedure described above, sample extracts were initially screened using eight sample concentrations along a two-fold dilution series that were assayed in triplicate. These sample concentrations ranged from 0.0078–1.0% of the total extract, corresponding to doses of ~0.7–1000 *Gambierdiscus* cells. The extract dissolution series was prepared in 100% MeOH, dried under high-purity N2 gas, redissolved in a minimum 100 μL of assay growth media, and 10 μL was added to assay well. Since a minimum of 100 μL was required to redissolve dried extracts, but only 10 μL of this could be used as a dose, a dosing range of 0.008–1.0% required using at least 10% of total sample extract. Due to the

nature of determining IC50 values from sigmoidal dose-response curves that meet quality control criteria, full quantitative N2a assays require the highest doses to result in <20% N2a cell viability and be 8 to 16-fold higher than the IC50 and at least 32-fold higher than doses that show no CTX-specific toxicity. Thus, samples in which a dosing of 1% of the total extract (requiring 10% of the extract to be used) demonstrated CTX-specific toxicity, but failed to cause <20% N2a cell viability, would not contain sufficient sample material for a quantitative assay result requiring a higher dosage. These low-CTX samples were considered to be positive detections of CTX, but their CTX content is described was "trace" rather than a numerical value. Samples for which the maximum dose of 1% sample extract or subsequent dilutions caused CTX-specific toxicity resulting in <20% N2a viability were analyzed by full quantitative assays. The dose for these quantitative assays was adjusted to achieve the maximum concentration at which non-O/V-treated N2a cells maintain >90% viability and O/V-treated cells have <20% N2a viability and generate a dose response curve meeting quality criteria. Quantitation was based on the mean sample IC50 values of six replicate dose response curves (measured in two assay plates, each containing triplicate dose response curves) that showed less than 15% variation.

#### 4.4.4. Tests of Matrix Effects and Sample Purification

In many samples, the doses that could be used for quantitation were limited by their level of non CTX-specific toxicity in N2a cells, i.e., the doses required to produce a toxin response in O/V-treated N2a cells also caused a significant toxic response in untreated N2a cells. This reflects our use of MeOH extracts that received no purification beyond partitioning with DCM and thus would contain a variety of cell metabolites, such as free fatty acids, that could be toxic to N2a cells at high doses. Possible interference of matrix compounds and additional sample clean-up were tested on four quantifiable samples representing the range of toxin concentrations across extracts in this study. Analyses of a dilution series for each of these four samples showed a linear, proportional response to dilution in quantifiable concentrations and only 7–16% variation in determined CTX concentration, indicating a lack of matrix effects on measured CTX-specific toxicity. Solidphase extraction (SPE) was performed on these samples using silica (Bond Elut Si; 100 mg; Agilent, Santa Clara, CA, USA). In all four samples, Si SPE purification caused a >40% reduction in toxicity to both O/V-treated and untreated N2a cells, indicating that CTX was being removed along with cytotoxic matrix compounds and that SPE clean-up would greatly affect accuracy of CTX measurement. Hence, further purification was not performed on any quantified samples to ensure accuracy rather than sensitivity. This decision was supported by C-CTX-1 spike recovery trials (below). The lowest concentration that could be quantified with confirmed CTX-specific toxicity was 0.28 ng C-CTX-1 eq. mL−<sup>1</sup> (in extract), which represents the effective limit of quantitation for the samples in this study.

To better determine the limits of detection and possibility of matrix interference for N2a analyses of natural epiphyte assemblages, a sample containing no detectable CTX activity (as determined by the screening procedure described above) was spiked with C-CTX-1 standard at 8 concentrations ranging from 0.01–0.5 ng C-CTX-1 mL−<sup>1</sup> (in extract) and the same range of concentrations were also produced in a dilution series (i.e., matrix concentration declined with CTX concentration). These tests indicated a limit of quantitation of 0.08 ng C-CTX-1 mL−<sup>1</sup> in unpurified algal extracts and that CTX quantitation was not affected by dilution of sample matrix. However, this limit is not directly comparable to the detection limit of non-spiked samples since C-CTX-1 is a major component of bioaccumulated CTXs in fish, but has not yet been attributed as a major component in the toxin profiles of *Gambierdiscus* or *Fukoyoa* [76].

#### *4.5. Statistical Analyses*

*Gambierdiscus* abundance and toxicity data were tested for normality and homoscedasticity using a Shapiro–Wilk test and a Brown–Forsythe test, respectively. Log-transformed abundance data was parametric and mean abundance between sites and years were compared with one-way ANOVA followed by Tukey's post hoc test for multiple comparisons. Toxicity data was highly skewed and non-parametric. A Box–Cox test was used to determine a transformation for toxicity data and all toxicity data was raised to the –2 power and mean values for sites and years were compared using a Welch's ANOVA and a Games– Howell test for multiple comparisons. All univariate statistical analyses were performed in R. Benthic community composition across sites and years were compared using a cluster analysis and non-metric multidimensional scaling and their multivariate similarity was measured with an ANOSIM test, all performed using Primer-E. Multivariate correlations between benthic community composition or environmental conditions and *Gambierdiscus* abundance or toxicity were examined in Primer-E using the BEST routine. All reported multivariate results had a significance level of 0.1% (*p* < 0.001). Significance in the PCA analyses were based on the broken stick criterion of Peres-Neto et al. [77]. Graphs were created using GraphPad Prism version 9.0.0 for macOS (GraphPad Software, San Diego, California USA, www.graphpad.com, accessed on 9 June 2021).

**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/10 .3390/toxins13060413/s1. Table S1: Pearson's correlations for environmental variables and the first two principal components from a principal components analysis of environmental data collected in this study. Table S2: Results of an ANOSIM test conducted in Primer-E, comparing macroalgal composition at Black Point, Coculus Rock, Flat Cay, and Seahorse Shoal in the U.S. Virgin Islands. Figure S1: A non-metric multidimensional scaling plot and results of ANOSIM test for benthic macroalgal composition at Black Point, Coculus Rock, Flat Cay, and Seahorse Shoal in the U.S. Virgin Islands. Figure S2: Examples of N2a assay dose-response curves for purified C-CTX-1 standards and corresponding representative field epiphyte samples.

**Author Contributions:** Conceptualization, A.R., T.B.S., M.L.R., D.M.A.; methodology, A.R., J.D.L., M.L.R., T.B.S., Y.X.; validation, J.D.L., A.R.; formal analysis, J.D.L., A.R., J.L.D.; investigation, J.D.L., M.L.R., T.B.S., Y.X., A.R.; resources, A.R., T.B.S., D.M.A.; data curation, J.D.L., T.B.S., J.L.D., A.R.; writing—original draft preparation, J.D.L., A.R., M.L.R., T.B.S., J.L.D.; writing—review and editing, T.B.S., D.M.A., Y.X.; visualization, J.D.L., J.L.D., A.R.; interpretation, J.D.L., M.L.R., T.B.S., J.L.D., A.R.; supervision, A.R., T.B.S., M.L.R., D.M.A.; project administration, A.R.; funding acquisition, A.R., T.B.S., M.L.R., D.M.A. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was funded in part by the National Oceanic and Atmospheric Administration, Ecology and Oceanography of Harmful Algal Blooms Program (ECOHAB publication number 984) through the CiguaHAB project (NA11NOS4780028), and also contributes to CIGUATOX (NA17NOS4780181) granted to coauthors AR, TBS, DMA, and MLR. Additional support was provided by NSF Partnerships in International Research and Education (1743802), and the Greater Caribbean Center for Ciguatera Research (NIH 1P01ES028949-01 and NSF 1841811). Financial support of YX was from the National Natural Science Foundation of China (41976155), the Natural Science Foundation of Guangxi Province (2020GXNSFDA297001).

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Complete environmental and physical data associated with this work is provided for open access as a downloadable file. Complete toxin and *Gambierdiscus* abundance data is presented in this manuscript but is also available on request to the corresponding author.

**Acknowledgments:** We greatly appreciate the extensive team of UVI graduate students and divers that contributed to the USVI coral reef monitoring program data and sample collection during the period of this study. Thanks to Robert Brewer, Sarah Heidmann, and Jonathan Jossart who analyzed benthic composition during this period. We also appreciate the efforts of Ana Garcia early in this project who assisted with sample extractions and preliminary toxicity analyses, and to Christopher Loeffler for early discussion on the cytotoxicity methodology employed.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


## *Article* **Experimental Evidence of Ciguatoxin Accumulation and Depuration in Carnivorous Lionfish**

**Isabel do Prado Leite 1,\*, Khalil Sdiri 2, Angus Taylor 3, Jérôme Viallon 4, Hela Ben Gharbia 5, Luiz Laureno Mafra Júnior 1,6, Peter Swarzenski 3, François Oberhaensli 3, Hélène Taiana Darius 4, Mireille Chinain <sup>4</sup> and Marie-Yasmine Dechraoui Bottein 2,\***


**Abstract:** Ciguatera poisoning is a food intoxication associated with the consumption of fish or shellfish contaminated, through trophic transfer, with ciguatoxins (CTXs). In this study, we developed an experimental model to assess the trophic transfer of CTXs from herbivorous parrotfish, *Chlorurus microrhinos*, to carnivorous lionfish, *Pterois volitans*. During a 6-week period, juvenile lionfish were fed naturally contaminated parrotfish fillets at a daily dose of 0.11 or 0.035 ng CTX3C equiv. g−1, as measured by the radioligand-receptor binding assay (r-RBA) or neuroblastoma cell-based assay (CBA-N2a), respectively. During an additional 6-week depuration period, the remaining fish were fed a CTX-free diet. Using r-RBA, no CTXs were detectable in muscular tissues, whereas CTXs were measured in the livers of two out of nine fish sampled during exposure, and in four out of eight fish sampled during depuration. Timepoint pooled liver samples, as analyzed by CBA-N2a, confirmed the accumulation of CTXs in liver tissues, reaching 0.89 ng CTX3C equiv. g−<sup>1</sup> after 41 days of exposure, followed by slow toxin elimination, with 0.37 ng CTX3C equiv. g−<sup>1</sup> measured after the 6-week depuration. These preliminary results, which need to be pursued in adult lionfish, strengthen our knowledge on CTX transfer and kinetics along the food web.

**Keywords:** ciguatoxins; experimental exposure; lionfish; trophic transfer; toxin accumulation; ciguatera poisoning

**Key Contribution:** In this study; ciguatoxins (CTXs) were detected in the liver of juvenile lionfish *Pterois volitans* fed naturally CTX-contaminated parrotfish *Chlorurus microrhinos* fillets. CTX remained in the livers of lionfish 43 days after the last exposure. Toxin levels in the muscular tissues were below the limits of detection of the receptor binding assay.

### **1. Introduction**

*Gambierdiscus* spp. are benthic dinoflagellates that may produce lipophilic ciguatoxins (CTXs), as well as other bioactive compounds, including maitotoxins (MTXs), gambierone, gambieroxides, gambierol, and gambieric acid [1–6]. Ciguatoxins are considered the

**Citation:** Leite, I.d.P.; Sdiri, K.; Taylor, A.; Viallon, J.; Gharbia, H.B.; Mafra Júnior, L.L.; Swarzenski, P.; Oberhaensli, F.; Darius, H.T.; Chinain, M.; et al. Experimental Evidence of Ciguatoxin Accumulation and Depuration in Carnivorous Lionfish. *Toxins* **2021**, *13*, 564. https://doi.org/10.3390/toxins 13080564

Received: 23 June 2021 Accepted: 3 August 2021 Published: 11 August 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

primary cause of Ciguatera poisoning (CP) in humans, affecting up to 500,000 fish consumers every year, including rare lethal cases [7–10]. Affected people may experience temporary and/or persistent/recurrent neurological effects resulting from CTXs binding to voltage-gated sodium channels in excitable tissues [11–15]. Ciguatera symptoms include the tingling of body extremities and/or cold allodynia and, in more severe cases, cardiovascular and respiratory insufficiency leading to coma and death [12]. Chronic neurological manifestations may persist for weeks, months, and in severe cases, years after the first exposure to CTXs [8,16,17]. Symptoms may vary geographically according to the dominant toxin profile in each region. Gastrointestinal disorders such as nausea, vomiting, abdominal pain, and diarrhea are predominantly manifested in the Caribbean region, where fish are contaminated by the so-called Caribbean ciguatoxins (C-CTXs). In contrast, in the Pacific and Indian Oceans, neurotoxic symptoms may predominate following the consumption of fish containing P-CTXs and I-CTXs, respectively. Neurotoxic symptoms may include hallucinations mostly among victims from the Indian Ocean [8].

The effects of CTXs on marine fauna are less documented. In fish, adverse effects from exposure to CTX have been observed in fresh/brackish water species such as blueheads, *Thalassoma bifasciatum* [18], juveniles of *Oreochromis* sp. [19], larvae and embryos of the genus *Oryzias* [20–24], and adult *Mugil cephalus* [25]. Ciguatoxins found in the brain, liver, and muscles of stranded *Monachus schauinslandi* monk seals [26] suggest that marine mammals may also suffer from CTX exposure, and that these compounds persist within the complex marine food webs. However, marine species experimentally fed with CTX (*Naso brevirostris* [27] and *Epinephelus coioides* [28]) did not seem sensitive to the effects of CTX. The fish resistance mechanism to CTX is still unknown.

Following the ingestion of *Gambierdiscus* cells by herbivorous and omnivorous organisms, CTX-like toxins are biotransformed and transmitted along the trophic food webs to top-chain carnivores, including fish, such as groupers and snappers, and marine mammals [28–31]. More than 400 species of fish are suspected as potential vectors of CP to humans [32]. Coral reef fish known to accumulate CTXs include barracuda, grouper, snapper, moray eel, parrotfish, trevally, and wrasse [33,34]. In general, higher-level carnivorous fish exhibit greater toxin concentrations than smaller fish and herbivores [35–37], but this is not always the case [38–42]. The complex processes and kinetics involving CTX accumulation, elimination, and trophic transfer to carnivores are still poorly understood and limited to field observations [42] and scarce laboratory studies [28].

Lionfish, *Pterois volitans* (Scorpaenidae), are carnivorous fish with venom-containing spines [43]. They are native to the Indo-Pacific region and were mostly likely introduced through aquarium releases into the Atlantic Ocean, becoming invasive to the Caribbean Sea and the Gulf of Mexico [44,45]. Due to the absence of natural predators for lionfish in reef ecosystems where they have been recently introduced, their use as a fishery resource has been stimulated to reduce the adverse ecological impacts on reef communities [46,47]. However, as with many other species, lionfish may accumulate CTXs and reach toxin concentrations above safety levels for human consumption as recommended by the U.S. Food and Drug Administration (0.1 ng C-CTX1 equiv. g−1, or 0.01 ng CTX1B equiv. g−1) [48]. Concentrations of CTXs in lionfish vary geographically, primarily depending on the local abundance and composition/dominance of *Gambierdiscus* species. In the Caribbean Ciguatera-endemic region, for instance, CTX accumulation in lionfish muscles can be highly frequent (≥40–50%), and reach levels up to 0.3 ng C-CTX1 equiv. g−<sup>1</sup> [46,49]. This suggests that *P. volitans* could become a common vector for CP in that region, where other top predators (e.g., snapper, barracuda, and grouper) can accumulate even higher CTX levels, posing a risk to human health [46,49]. Those toxin-equivalent concentrations must be considered carefully, however, because the venoms produced by lionfish itself can potentially interfere with the toxicity bioassays commonly used to quantify CTX. To avoid false-positive results on CTX tests, cooking the fillets of lionfish is recommended before testing for the presence of ciguatoxin. This procedure denatures the scorpaenitoxins, leaving only CTX-related toxicity if present [50]. Regardless, at least one case of CP following the consumption of lionfish has been already confirmed [51].

Considering the ecological relevance of *P. volitans* associated with its potential threats as an invasive species and emerging CP vector, this species was selected in the present study to study the transfer of CTX from a prey to a carnivorous reef fish predator. The potential effects of CTXs, as well as their transfer, accumulation, and elimination were investigated during a long-term (12-week) laboratory feeding experiment using naturally CTX-contaminated steephead parrotfish, *Chlorurus microrhinos* (formerly *Scarus gibbus*), fillets as source of toxins. The presence and concentration of CTXs were evaluated in liver and muscle tissues of juvenile *P. volitans* during both toxin uptake and depuration phases, using distinct analytical methods such as radioligand-receptor binding assay (r-RBA), neuroblastoma cell-based assay (CBA-N2a), and liquid chromatography coupled to tandem mass spectrometry (LC-MS/MS).

#### **2. Results and Discussion**

#### *2.1. Experimental Setting and Fish Behavior*

The experimental feeding encompassed a 12-week period, including 6 weeks of exposure to CTX and 6 weeks of depuration. A gradual decrease in feeding activity was observed over the course of the experiment in both CTX exposed and control fish, although the propensity was greater among exposed fish (Figure 1). Increased food rejection episodes were indeed registered, mainly after the fifth week of experiment, during the late exposure period. The proportion of non-eating fish was 62% in the group exposed to CTX and 32% in the control group at late exposure phase (Figure 1). Such changes in feeding behavior were not observed in previous experiments using herbivorous fish (*Naso brevirostris*) or carnivorous groupers (*Epinephelus coioides*) exposed to diets containing 0.4 ng CTX3C equiv. g−<sup>1</sup> over 16 weeks and ~1.0 ng CTXs equiv. g−<sup>1</sup> day−<sup>1</sup> (mean of CTX1B, 52-*epi*-54-deoxyCTX1B, 54-deoxyCTX1B quantifications) over 30 days of exposure, respectively [27,28]. However, the food rejection episodes observed during the present experiment did not seem to affect the CTX uptake and accumulation, since toxin levels were, in fact, detected in fish liver at the final exposure stage due to prolonged feeding (4–5 weeks) on CTX-containing parrotfish flesh (Table 1).

**Figure 1.** Proportion of control and CTX-exposed lionfish rejecting the food in relation to the number of remaining individuals at a given time during both the exposure and depuration periods.


**Table 1.** Detection and quantification of ciguatoxins in individual lionfish livers during the experimental exposure and depuration periods by r-RBA analysis.

\*\* LOD = the limit of detection of the r-RBA was 0.75 ng CTX3C equiv. g−1.

Although laboratory conditions appeared to be optimal and constant throughout the experiment, unexpected deaths were registered among both control and exposed lionfish (15 out of 48 individuals in total). Mortality also affected individuals belonging to a separate batch received after the beginning of the experiment, maintained under similar conditions in the facility for future experiments. The specimens used in this study may have become more sensitive and prone to diseases such as bacterial infections when transferred to a laboratory setting. Consequently, these unexpected fish mortalities led to the loss of experimental replicates. Despite the reduction of replicates and considering that (i) fish showed no signs of suffering during the experiment, (ii) CTX-contaminated matrices were difficult to obtain, and (iii) published information of CTX trophic transfer under controlled conditions are still very limited, we decided to maintain the experimental protocol. Diet comparisons (i.e., exposed vs. control group) were thus achieved using randomly sampled individual fish as replicates.

Increased food rejection episodes may have contributed to the fish deaths observed during our experiment, although this could not be confirmed. Indiscriminate, sporadic mortality episodes registered during the experiment seemed unrelated to the exposure to CTX. Out of the 32 CTX-dosed fish, 11 fish died, mostly on week 7 (after 1 week of depuration). Similarly, 4 out of 16 control fish died, but in this case, most deaths were registered on week 12, at the end of the experiment. The cumulative proportion of fish mortalities in the exposed group (36%) was similar to that of the control (28%).

During the entire experiment, exposed lionfish exhibited no clear signs of acute intoxication, such as erratic swimming, rapid gill movement, loss of equilibrium, and fin paralysis, as described in CTX-exposed mullet *Mugil cephalus* [25]. Likewise, in natural environments, CTX-contaminated fishes, including lionfish in the Caribbean region [49], as well as snapper and grouper on Pacific coral reefs [42], have been found to exhibit no signs of intoxication. Thus, coral reef fish species facing repeated exposure to CTXs in the natural environment may have developed resistance mechanisms to these compounds, perhaps binding them to particular soluble proteins in the skeletal muscle [52,53], or storing them in compartments where they would be biologically unavailable.

#### *2.2. Toxin Levels in the Food and Fish Dosing*

The average CTX content in the naturally contaminated parrotfish (*Chlorurus microrhinos*) fillet homogenate (toxic food) was estimated as 2.27 ± 0.4 ng CTX3C equiv. g−<sup>1</sup> (coefficient of variation, CV = 17.6%) by r-RBA, with no significant difference (*p* < 0.05) among the eight subsamples tested. Analysis by CBA-N2a confirmed the presence of CTXs in the CTX-contaminated food, estimated at a concentration of 0.70 ± 0.02 ng CTX3C equiv. g−<sup>1</sup> (*n* = 3) with CV of 2.8% (Supplementary Materials, Figure S1). Unfortunately, the toxin levels in parrotfish flesh were too close to the LOD and LOQ values by LC-MS/MS (up to 1.5 ng CTX3C equiv. g−<sup>1</sup> and 4.7 ng CTX3C equiv. g−<sup>1</sup> of flesh, respectively), to allow unambiguous determination of CTXs using this technique. As expected, the CTX-free food (control) exhibited no activity by r-RBA analysis in our experimental conditions.

Given the toxin content in the CTX-contaminated food and the daily ration of 0.05 g food per g of fish, individuals from the CTX-exposed group received a dose equivalent to 0.11 ng CTX3C equiv. g−<sup>1</sup> d−<sup>1</sup> according to r-RBA estimations, or 0.035 ng CTX3C equiv. g−<sup>1</sup> d−<sup>1</sup> by CBA-N2a quantification. The dose administrated herein can be representative of an exposure that can eventually occur in the reef environment (i.e., environmentally relevant dose), as high levels of CTXs (up to 10.7 ng CTX3C equiv. g<sup>−</sup>1) have been reported in parrotfish (*C. microrhinos*) sampled from French Polynesia islands [39].

Considering the average amount of food supplied and the toxin content in the experimental diet, the total toxin ingested per lionfish (initial body weight: 15.8 ± 9.6 g; *n* = 32) was estimated by r-RBA as 1.74 ng CTX3C equiv. day−1, totalizing 52.2 ng CTX3C equiv. and 71.3 ng CTX3C equiv. over 30 and 41 days, respectively, assuming that the fish accept all food portions offered. Based on the CBA-N2a analysis, each exposed lionfish may have ingested, on average, a 3.2-fold lower dose.

#### *2.3. Toxin Levels in Fish Tissues*

CTX were detectable by r-RBA in the liver of fish, yet not consistently. When quantified by this assay, toxin concentrations in the livers of exposed lionfish reached 9.4 ng CTX3C equiv. g−<sup>1</sup> at day 41 of exposure to CTX-contaminated food. The highest toxin concentration (9.77 ng CTX3C equiv. g<sup>−</sup>1) was measured at late depuration phase (Table 1). As expected, CTX was not detected in the livers of control fish using either r-RBA (LOD: 0.75 ng CTX3C equiv. g−1) or LC-MS/MS (LOD: up to 1.14 CTX3C equiv. g−<sup>1</sup> of liver). In fact, this latter analytical method proved to be the least sensitive one. Unfortunately, the limited sample volumes remaining from other technique, as well as the relatively low concentrations of individual CTX compounds present in our samples, together with the matrix effect, did not allow us to make any further clarification on the toxin profile of CTXexposed fish. In French Polynesia, CTX composition of steephead parrotfish *C. microrhinos* typically includes CTX3C type compounds (71.1% as CTX3C and its M-seco form) and, to a lesser extent, CTX4A type (28.8% as CTX4A and its M-seco form) [33,54]. However, CTX profiles in fish are species-specific and can differ from regional variations [55,56]. The typical toxin profile reported in *Gambierdiscus* species found in Pacific coral reef is also mostly composed of the less polar CTXs (48% CTX3C, 34% 49-epiCTX3C and 13% CTX4A) [33].

Following CTX absorption, toxin metabolization in fish may affect the depuration and elimination process [37,55]. Of note, as suggested in the present study for lionfish, low CTX depuration rates were also reported in other carnivorous fishes, including experimentally exposed orange-spotted grouper (*Epinephelus coioides*) [28], as well as wild-caught moray eels *Gymnothorax javanicus* (*Lycodontis javanicus*) [57] and red snapper (*Lutjanus bohar*) [58]. For the latter species, the fish would require up to 30 months to completely eliminate the acquired toxin load. Moreover, the livers of contaminated orange-spotted groupers showed faster elimination rates for CTX1B, 52-*epi*-54-deoxyCTX1B, and 54-deoxyCTX1B, which are already highly metabolized toxins, compared to other tissues such as skin, gills, and muscles, indicating different CTX elimination rates among tissues of exposed fish over 30 days [28].

Pooled liver extracts from control fish caused no cytotoxic effects on N2a cells in either OV<sup>−</sup> or OV+ conditions throughout the experiment. Conversely, sigmoidal doseresponse curves with a negative slope were obtained under the OV<sup>+</sup> condition for liver extracts from fish sampled after 30 and 41 days of exposure to toxic food (Figure 2), and from those sampled after a depuration period of 8, 29, and 43 days (Figure 3). Based on the CTX-like composite toxicity using the CBA-N2a results, toxin concentrations in pooled lionfish livers were estimated as 1.36 ng CTX3C equiv. g−<sup>1</sup> and 0.89 ng CTX3C equiv. g−<sup>1</sup> after 30 and 41 days of exposure to toxin-containing food (herbivore *C. microrhinos* parrotfish fillets). Toxin concentration values decreased gradually over the depuration period, reaching 0.81 ng CTX3C equiv. g<sup>−</sup>1, 0.52 ng CTX3C equiv. g−1, and 0.37 ng CTX3C equiv. g−<sup>1</sup> in pooled extracts from 8, 29, and 43 days of depuration on non-toxic food, respectively (Table 2). Considering the decrease in CTX-like cytotoxicity corresponding to a 0.44 ng CTX3C equiv. g−<sup>1</sup> difference between days 8 and 43 of depuration (see Table 2), it would take an additional 35-day depuration period for lionfish livers to reach the CBA-N2a LOD of 0.06 ng CTX3C equiv. g<sup>−</sup>1. The kinetics of CTX elimination in lionfish liver, however, should be examined with caution, considering: (i) the limited number of time points for sampling during depuration stage (8, 23, and 43 days after last exposure), (ii) the use of pooled samples in CBA-N2a due to an insufficient amount of remaining individual tissues, and (iii) the limited amount of pooled tissues available, which did not allow for replication of the quantification of CTX in livers by CBA-N2a.

Livers of lionfish belonging to the CTX-exposed group weighed, on average, 0.64 g (3.4% of the total fish body weight), resulting in a toxin burden of up to 4.53 ng CTX3C equiv. and 6.04 ng CTX3C equiv. in this organ after 30 and 41 days of exposure, respectively (based on the r-RBA results). When referring to the CBA-N2a results, the average toxin burden in pooled livers ranged from 0.87 ng CTX3C equiv. to 1.90 ng CTX3C equiv. after 30 and 41 days of exposure, respectively. This represents a retention of 8.6% (r-RBA) or 6.8% (CBA-N2a) of the ingested toxin burden at the end of the exposure period. Recent studies have reported that carnivorous grouper (*Epinephelus coioides*) fish, fed over 30 days, demonstrated a toxin burden of ~1 ng CTX1B g−<sup>1</sup> d<sup>−</sup>1—a greater toxin dose than that used in our experiment (0.11 ng CTX3C equiv. g−<sup>1</sup> d−<sup>1</sup> or 0.035 ng CTX3C equiv. g−<sup>1</sup> d−<sup>1</sup> from r-RBA or CBA-N2a analysis)—and accumulated up to 25% of the ingested toxin load in the livers and 10% in muscles [28]. In contrast, after 16 weeks of dietary exposure to 0.4 ng CTX3C equiv. g−<sup>1</sup> d<sup>−</sup>1, juvenile *Naso brevirostris* accumulated ~2% of the exposed CTX3C amounts in muscular tissue [27].

In the present study, toxin concentrations in the muscles of lionfish belonging to the dosed group were always below the limit of detection, as determined by r-RBA. The sample amounts remaining from this first analysis were deemed insufficient for a supplementary evaluation by CBA-N2a. In natural environments, lionfish muscles were found to contain relatively high CTX levels (≥0.24 ng C-CTX1 equiv. g−1), indicating the potential of this species to cause human poisoning events [46,51]. When experimentally exposed to fourfold higher weight-specific daily doses of CTXs than those administered in our experiment, the herbivorous coral reef fish *Naso brevirostris* accumulated up to 3.24 ± 0.59 ng CTX3C equiv. g−<sup>1</sup> in muscular tissues [27]. This value is within the range usually found among naturally contaminated fish in French Polynesia [41]. Moreover, in CTX-exposed *N. brevirostris*, the total amount of toxins in muscles increased linearly over 16 weeks of toxin exposure [27]. The muscles of *N. brevirostris* quickly eliminated the incorporated ciguatoxins, contrary to what has been observed for mullets (*Mugil cephalus*) after nine toxic feedings with gel food containing *Gambierdiscus* cells [25,27]. Finally, muscular tissues of groupers (*Epinephelus coioides*) incorporated the equivalent of 0.34 ng CTX1B g−<sup>1</sup> d−<sup>1</sup> after 30 days of exposure to 1 ng CTX1B g−<sup>1</sup> d−<sup>1</sup> [28]. In the Caribbean region, the highest C-CTX1 levels detected in the muscles of wild-caught fish were 0.24 ng g−<sup>1</sup> in grey snapper (*Lutjanus griseus*), 0.3 ng g−<sup>1</sup> C-CTX1 equivalents in lionfish (*P. volitans*), 0.9 ng g−<sup>1</sup> in grouper (Serranidae), 13.8 ng g−<sup>1</sup> in black jack (*Caranx lugubris*), and 49 ng g−<sup>1</sup> in barracuda (*Sphyraena barracuda*) [49,59–61].

**Figure 2.** Quantification of CTX in fish liver during the exposure phase of the experiment. Doseresponse curves of N2a cells when exposed to increasing concentrations of fish liver extracts, obtained from the exposure phase of the experiment, in OV<sup>−</sup> (open symbols) and OV<sup>+</sup> (solid symbols) conditions; (**A**) CTX3C (◆), (**B**) non-exposed control fish at day 42 (pool of 3 specimens) (Δ/-), (**C**) exposed fish at day 30 (pool of 4 specimens) (-/•), and (**D**) exposed fish at day 41 (pool of 5 specimens) (/). Data represent the mean ± SD of 1 assay, with each point run in triplicate. The dotted vertical line corresponds to the maximum concentration of liver tissue (MCE) that does not induce a matrix effect on the assay, which was estimated at 20 mg mL−<sup>1</sup> of fish liver extracts.

**Figure 3.** Quantification of CTX in fish liver during the depuration phase of the experiment. Doseresponse curves of N2a cells when exposed to increasing concentrations of pooled fish liver extracts, obtained from the depuration stage of the experiment, in OV<sup>−</sup> (open symbols) and OV<sup>+</sup> (solid symbols); (**A**) 8 days of depuration (pool of 3 specimens) (-/•), (**B**) 29 days of depuration (pool of three specimens) (/), and (**C**) 43 days of depuration (pool of two specimens) (Δ/-). Data represent the mean ± SD of one assay, with each point run in triplicate. The dotted vertical line corresponds to the maximum concentration of liver tissue (MCE) that does not induce the matrix effect, which was established at 20 mg mL−<sup>1</sup> of fish liver extracts.

Despite the loss of replication and the need for a longer period of experimentation, our results indicate CTX transfer from contaminated prey to carnivorous *P. volitans*. Toxins concentrated into lionfish livers, while their transport to the muscles could not be reliably assessed. In other fish species, CTXs appear to be primarily accumulated in the liver, being secondarily distributed to the muscles and other tissues as the concentrations increase [28].


**Table 2.** Estimation of the EC50 values and the CTX-like concentration of the pooled lionfish liver samples, as determined by the CBA-N2a analysis of the control (CTX-free food) and during exposure (CTX-contaminated food) and depuration (CTX-free food) periods of the experiment.

f\* ND = not determined as control samples gave no cytotoxicity at the MCE in both OV conditions. \*\* LOD = limit of detection of the CBA-N2a (0.06 ± 0.01 ng CTX3C equiv. g−1).

#### **3. Conclusions**

The accumulation of CTX in the livers of juvenile lionfish *Pterois volitans* became evident after 5 weeks of feeding toxin-contained parrotfish *Chlorurus microrhinos* fillets. The toxin levels in the muscles of lionfish were always below the limits of detection, as determined by r-RBA, suggesting potential differential tissue distribution over the experiment. However, this could not be confirmed by a more sensitive method, such as CBA-N2a due to the insufficient amount of flesh. Lionfish retained detectable toxin levels in their livers at 43 days after the last exposure, indicating a slow elimination process. No acute intoxication signs were observed throughout the experiment, suggesting the resistance of *P. volitans* to CTXs.

#### **4. Materials and Methods**

#### *4.1. Fish Acclimation and Maintenance*

Juvenile lionfish, *Pterois volitans* (Scorpaenidae), originating from Bali, Indonesia, were acquired from a supplier (Tropic Nguyen, France). The juvenile lionfish were selected based on maintenance and individual replication needs considering the laboratory setting and aquarium size. Upon reception, the lionfish were acclimated for approximately 2 weeks in a 2000 L open-circuit tank. The thank was filled with flowing 1 μm filtered seawater at 400 L h−1, 25 ± 0.5 ◦C, pH 8.1 ± 0.1, and a salinity of 39, and maintained under a 12:12 h (light:dark) cycle and permanent aeration. Fish were fed once a day during acclimation in amounts higher than needed for maintenance (equivalent to 5% of their body weight). The food consisted initially of living prey (guppy or seabream). Then, inert frozen food (*Antherina boyeri* and krill) was gradually incorporated into the diet before the final transition to the frozen, CTX-containing parrotfish (*Chlorurus microrhinos*) fillet.

#### *4.2. Preparation of the Experimental Diets*

The toxic material used for the feeding experiments consisted of naturally CTXcontaminated fillets of the steephead parrotfish *Chlorurus microrhinos* collected from Moruroa Atoll (Tuamotu Archipelago, French Polynesia). Parrotfish were selected based on the CTX levels in their fillets. Then, 7.8 kg (wet weight) of selected fillet were homogenized using an industrial mixer to obtain a toxic, homogenous fish fillet matrix. Toxicity of the prepared food was evaluated by means of cytotoxicity using CBA-N2a and binding affinity using r-RBA composed of 3 × 10 g and 8 × 5 g aliquots of the dried homogenate, respectively.

Different food preparations using Gelly Belly™ (Gel Food, Florida Aqua Farms, Inc., Dade City, FL, USA) and frozen food were tested. Gelly Belly™ consists of a gelatin-based food mixed with microalgae, seaweeds, fish, and krill meal, with added vitamins and minerals. The most appropriate food presentation for lionfish proved to be homogenized

cube-shaped frozen fillets. Frozen food dices were individually offered to lionfish using forceps, which were instantaneously caught and ingested without significant particle loss.

The food provided to the individuals in the control group—and to all remaining lionfish during the depuration phase—was composed of homogenized fillet of farmed seabream (*Sparus auratus*; ~120–140 g total body weight), originated from Turkey and supplied by Relais d'Or (France). After homogenization, seabream fillets were fashioned into cubes, placed in plastic bags, and kept at −18 ◦C until being offered to the carnivorous fish as frozen cubes. A sample of seabream fillets was collected and further tested for CTX presence.

#### *4.3. Experimental Design*

After acclimation, *P. volitans* individuals (*n* = 48; 5.05–34.0 g initial wet weight) were distributed among seven aquaria, allocating six fish per tank, except in two tanks where ten and eight fish were placed due to the smaller size of some individuals (~5 g) (Figure 4). This step aimed at obtaining similar fish sizes and total living biomass (15.2 g ± 9.52, mean ± SE) across the experimental tanks. Tanks were randomly assigned to either toxin-containing (5 tanks) or control diet (2 tanks), each containing 100 L of filtered seawater maintained under constant aeration and flow (100 L h−1), at the same conditions of temperature, salinity, and pH used during the acclimation phase. To reduce animal stress, three tubes of 10 cm × 20 cm (D × L), assembled as a pyramidal structure, were placed inside the aquaria to provide shelter for the fish (Figure 4). Finally, the aquarium walls were covered with opaque plastic coating to limit potential stress caused by the human presence in the laboratory. Standards of animal welfare were rigorously maintained throughout the experiment. All procedures were carefully conducted to minimize handling and reduce physiological stress.

**Figure 4.** Design of the experiment, with 5 tanks initially containing 6 or 8 fish (*n*) assigned to the CTX-exposed treatment (T) and 2 tanks for control (C) containing 6 fish and 10 fish (*n*). The image represents a picture of the juvenile lionfish (*Pterois volitans*) individuals and their shelters made of dark tubes in one of the experimental tanks.

After 6 weeks of exposure to either the toxin-containing or control diet, fish entered the depuration stage, during which they received non-contaminated frozen food (homogenized seabream fillets) for an additional 6-week period. Fish feedings were carefully conducted throughout the experiment, assuring quick capture (usually <5 min) and no significant particle rejection. Some individuals in the control group were maintained over the entire duration of the experiment (12 weeks) and served as reference for possible visual

behavior alterations and/or intoxication symptoms relative to dosed fish. The behavioral pattern was observed daily in all tanks during the feeding period throughout the 12-week experiment. During the exposure and depuration phases, the number of deaths was computed in relation to the total number of fishes at the beginning of the experiment, while the percentage of fish rejecting the toxic food considered the average number of individuals remaining in each experimental condition.

#### Feeding Experiment

During the experiment, all lionfish individuals were fed in a constant proportion to 4–5% of their body weight per day, as recommended for juvenile fish. The amount of food supplied was adjusted weekly based on the average fish weight for each tank. The lionfish were individually fed, always in the afternoon (2:00 p.m.), 5 days a week. Individuals were observed over 30 min to assure complete food ingestion after every feeding procedure. At the end of this time interval, any rejected food was removed from the tank during the daily cleaning procedure. Lionfish usually consumed the entire portion of toxic food offered in less than 15 min. In addition, the tanks were connected to an open system under a constant seawater flow. Thus, any residual toxin eventually released from the food—or excreted by the fish—was quickly eliminated from the tanks. Therefore, the source of CTX to the exposed fish was considered limited to the food pathway.

After 30 days of exposure, four CTX-exposed and three control lionfish were randomly collected from the control tanks. Likewise, between two and five individuals from the exposed group were sampled each time, at the end of the exposure period (41 days) and after 8, 29, and 43 days of the following depuration stage. Finally, three additional control individuals were collected after 8, 29, and 43 days of the depuration stage (Table 3). Fish euthanasia was achieved using an overdose of eugenol, and the death was confirmed by the absence of respiratory movements [62]. Lionfish were dissected immediately following death to ensure the integrity of collected tissues. For toxin analysis, muscle and liver samples were individually weighed (wet weight) and stored in plastic tubes at −18 ◦C.

**Table 3.** Number of CTX-exposed juvenile *Pterois volitans* individuals (*n*) sampled after a given exposure or depuration period. In addition, fish fed non-toxic food (control) were sacrificed after 30 days of exposure and after 8, 29, and 43 days of depuration period (3 individuals at each time).


*4.4. Toxin Determination in Food and Exposed Fish*

#### 4.4.1. Sample Extraction

The extraction of CTXs from the muscle and liver samples followed the procedures described in previous studies [27,63,64]. Briefly, each tissue sample from individual lionfish was extracted as a whole when the tissues weighed <4 g, or partially, in 4 g aliquots after homogenization with T-25 digital Ultra Turrax (IKA Works, Staufen, Germany). Tissue samples were then cooked in Falcon tubes in a water bath at 70 ◦C for 15 min, homogenized in acetone (3 mL g−<sup>1</sup> tissue) using a sonication probe (Branson digital probe cell breaker) for 2 min at 30% duty, and centrifuged for 3 min at 1400× *g*. The supernatant was recovered from the tubes, and the tissue pellet was homogenized twice again in acetone, as previously described. After three successive extraction steps, the supernatant fractions were combined and evaporated under nitrogen gas flow (Turbovap) in a water bath at 60 ◦C. Dried

extracts were resuspended in 5 mL of aqueous methanol (MeOH/H2O 90:10), and the lipids were removed by solvent–solvent separation (three times) after the addition of an equal volume of *n*-hexane. The 1:1 solvent–solvent fraction was separated (3×) into an aqueous phase (MeOH/H2O 60:40) and dichloromethane (DCM). Finally, the organic phase (DCM), containing the CTXs, was evaporated with nitrogen, resuspended in pure MeOH to 10 g tissue equivalent (TE) mL−1, and stored at −<sup>18</sup> ◦C until further analysis. Frozen homogenized fillet cubes of parrotfish *Chlorurus microrhinos* (toxic food) and seabream (control food) were extracted following the same procedure described for lionfish tissues, adjusting the sonication time to 20 min to ensure complete fish cell lysis.

#### 4.4.2. Toxin Analysis

Radioligand-Receptor Binding Assay (r-RBA)

The presence and quantification of CTXs in parrotfish (*C. microrhinos*) fillet homogenates (toxic food) and in selected extracts of fish muscle and liver were first determined using the r-RBA [65]. This detection method is based on the binding competition between CTXs from the sample and a radiolabeled brevetoxin (tritiated PbTx-3) for their common receptor on the voltage gated sodium channel (Nav)—using a porcine brain homogenate (Sigma Aldrich, St. Louis, MO, USA) membrane preparation [11,65].

The assay was performed on a microplate according to the method reported by the authors of [63], modified according to previous studies [27,64,65]. First, 35 μL of phosphatebuffered saline (PBS-Tween®) with bovine serum albumin (BSA) (1 g L−1) was added to each well of a 96-well filtration microplate (MultiScreen HTS FB Filter Plate MSFBN6B50, Millipore) to moisturize the membrane filter. Subsequently, 35 μL of either the standard CTX3C provided in dried form by Wako-Pure Chemicals, Osaka, Japan (2.85 × <sup>10</sup>−<sup>9</sup> to 1.92 × <sup>10</sup>−<sup>12</sup> M), a solution of PbTx-3 (Latoxan, Rosam, France) (1.8 × <sup>10</sup>−<sup>8</sup> M) used as an internal assay Quality Control (QC) (3 × <sup>10</sup>−<sup>9</sup> M in assay), or the diluted samples were added to each corresponding well, after vortex-mixing and sonication. Likewise, only one concentration of fish sample was tested, i.e., 0.6 g tissue equiv. mL−<sup>1</sup> for the parrotfish toxic food, and 0.6 g tissue equiv. mL−<sup>1</sup> and 0.12 g tissue equiv. mL−<sup>1</sup> for lionfish muscle and liver, respectively. Each sample was tested in duplicate wells in one experiment. Then, 35 μL of the working solution of [3H] PbTx-3 (1 nM assay concentration) and 195 μL of the diluted brain membrane homogenate (0.8 mg protein mL<sup>−</sup>1) were sequentially added to all wells. After incubation for 1 h at 4 ◦C, each well was washed (3×) with 200 μL of ice-cold phosphate-buffered saline solution (PBST) and filtered using a MultiScreen HTS vacuum collector system (Milipore, Billerica, Massachusetts, USA) to remove the excess radiotracer [ 3H] PbTx-3. Only the membrane receptor-bound toxin molecules were retained in the filter at this stage. After filtration, the microplate was placed on a counting cassette (Perkin-Elmer rigid 96 plate 14105), with 50 μL of liquid scintillant (Optiphase, Perkin-Elmer, USA) per well. Finally, the plate was incubated in the dark at room temperature for 2 h prior to quantification of the radioactivity in a beta-microplate counter (MicroBeta2, Perkin-Elmer, Waltham, MA, USA).

#### Liquid Chromatography Coupled with Tandem Mass Spectrometry (LC-MS/MS)

Analysis by LC-MS/MS was performed on liver samples of lionfish and in CTXcontaminated parrotfish *Chlorurus microrhinos* homogenates to confirm the presence of CTXs in the extracts. Aliquots of the extracts remaining from the r-RBA analysis (10–100 μL) were dried off with nitrogen gas and resuspended with the following volume of 90% aqueous MeOH: 200 μL for lionfish liver samples and 500 μL for parrotfish homogenate, yielding a final concentration of 0.5–2.6 g liver equiv. mL−<sup>1</sup> and 0.4–1.1 g flesh equiv. mL−1, respectively. Different dilutions reflected the amount of matrix available in each case. Toxin determination was performed on a UHPLC system (UFLC Nexera, SHI-MADZU, Japan) coupled to a hybrid triple quadrupole-linear ion-trap API4000 Qtrap mass spectrometer (ABSciex®, Framingham, MA, USA), equipped with a TurboV® electrospray ionization source.

Eluents consisted of deionized water (A) and 95% acetonitrile (B), both containing 2 mM ammonium formate and 50 mM formic acid. The following linear elution gradient was run at 0.4 mL min−<sup>1</sup> through a Zorbax Eclipse Plus C18 column (50 × 2.1 mm, 1.8 μm, 95 Å; Agilent Technologies, Santa Clara, CA, USA), maintained at 40 ◦C: 78 to 88% B in 10 min, held at 88% B for 4 min, decreased back to 78% in 1 min, and held during 5 min to equilibrate. Samples were kept at 4 ◦C during the analysis, and 5 μL aliquots were injected into the system. Scheduled MRM scanning (90 s detection window; 2 s target scan time) was applied in positive electron spray ionization (ESI+) mode, using the following optimized parameters: curtain gas at 25 psi; ion spray at 5500 V; turbo gas temperature at 300 ◦C; gas 1 and 2 at 40 and 60 psi, respectively; declustering potential at 105 V; and entrance potential at 10 V. A list of MRM transitions (*m/z*) scanned in ESI+ for the detection of CTX-like compounds is given in Table 4, along with the respective collision energy (CE) and targeted retention time values. Instrument control, data processing, and analysis were conducted using Analyst software 1.6.2 (SCIEX, CA, USA). Due to the lack of standards, calibration curves of CTX3C (Wako, Japan) were applied to calculate the concentration of every compound detected, assuming an equivalent molar response for the other analogs.

**Table 4.** List of MRM transitions (*m/z*) used in ESI+ to detect CTXs by LC-MS/MS.


Cell Based Assay on Neuroblastoma (CBA-N2a)

The neuroblastoma (N2a) CCL 131 cell line (ATCC) was used in this study, and the CBA-N2a was performed following the protocol described by the authors of [66]. Briefly, the 60 inner wells of several 96-well microplates were seeded with 200 μL of a 5% FBS culture medium at an initial cell density of 50,000 ± 10,000 cells well−<sup>1</sup> (exponential growth phase) and left to grow for 26 h in an incubator at 37 ◦C and 5% CO2. Among the several microplates run in parallel, one microplate served as the Reference Cell Viability (RCV) control to establish the initial cell viability of N2a cells using an MTT assay [66]. For each microplate, the mean absorbance of DMSO control (12 outer wells filled with DMSO only) was subtracted from each raw absorbance value (60 inner wells containing N2a cells), and all viability data were expressed in net absorbance data.

The remaining microplates were treated as follows. The growth medium was renewed by the addition of 200 μL of 2% FBS culture medium with 90 μM of ouabain (O) and 9 μM of veratridine (V) for non-destructive treatment in the OV+ condition (bottom half of the microplate). In the OV− condition, the growth medium was renewed by the addition of 200 μL of 2% FBS culture medium (upper half of the microplate). Appropriate controls in both OV conditions, namely COV<sup>−</sup> and COV+, were established by adding 10 μL of 2% FBS culture medium to verify the final cell viability and the non-cytotoxicity of the OV treatment, respectively, in the absence of CTXs. The implementation of additional quality check controls (QC) to both OV- and OV<sup>+</sup> conditions, namely QCOV<sup>−</sup> and QCOV+, was also undertaken to check for the specific effect of voltage gated sodium channel (VGSC) activators. Basically, 10 μL of PbTx3 (Latoxan, France) at 0.1 μg mL−<sup>1</sup> were added in triplicate wells in both OV conditions to reach a final concentration of 4760 pg PbTx3 mL−<sup>1</sup> in the wells. As CTX3C was tested in parallel with the fish samples, eight-point 1:2 serial dilutions of the CTX3C standard (Institut Louis Malardé) and sample stock solutions were prepared (100 μL per concentration) using a U-bottom 96-well microtiter. Then, 10 μL of CTX3C concentrations were directly added in triplicate wells under the non-destructive OV<sup>+</sup> condition (85.7/8.57 μM final concentrations) and not under OV- condition, since no cytotoxicity occurs in the absence of the O/V treatment [66]. The final concentrations of CTX3C tested ranged from 0.15 to 19.05 pg mL−1. The three parrotfish (toxic food) aliquots were tested with final concentrations ranging from 74 to 9524 ng mL−<sup>1</sup> of dry extract (corresponding to 0.23–29.76 mg flesh equiv. mL−1). Each concentration was tested in triplicate wells in both OV conditions, in three independent microplates run the same day.

Liver extracts were further analyzed by CBA-N2a. To this end, all liver extracts prepared from individual fish collected from a single tank at the same sampling interval (days 30, 41 of the exposure phase and days 8, 29, and 43 of the depuration phase). Pooled samples ranged from 0.11–0.51 g fresh weight equivalent of fish liver. These samples were resuspended in a solution of 5 μL MeOH, 15 μL dimethyl sulfoxide (DMSO), and 2% fetal bovine serum (FBS) RPMI medium, providing concentrations of sample stock solutions ranging from 0.442 g fresh liver equiv. mL−<sup>1</sup> to 1.567 g fresh liver equiv. mL−1, with no more than 10% of solvent used during the serially two-fold dilution procedure. As dry extract weights were not available for lionfish livers, fresh fish liver equivalents could only be considered to estimate the concentration ranges tested. Pools of lionfish liver from the exposure experiment were tested together in one CBA-N2a experiment. In the same way, pooled liver from the depuration phase were also tested together in another CBA-N2a experiment. For each CBA-N2a experiment, CTX3C was tested in parallel with the lionfish liver samples, each concentration tested in triplicate wells in both OV conditions. Full dose-response curves were used for accurate CTX quantification. These two CBA-N2a experiments could not be repeated due to the very small amount of extract available. In all microplates, peripheral wells received 200 μL of sterile distilled water, and microplates were left to incubate overnight for about 19 h prior to the determination of the final cell viability using the MTT assay with an incubation time of 45 min.

## *4.5. Data Analysis*

#### 4.5.1. r-RBA

The r-RBA result of each extract was determined after the assay performance was confirmed by the Quality Control (QC), together with the additional parameters of the CTX3C standard curve (Supplementary Materials, Figure S2), such as the effective concentration inducing a 50% values effect (EC50), and the slope of the linear portion of the standard curve. Sample CTX quantification was only completed when the CPM measurement of a sample dilution fell on the linear part of the CTX3C standard curve and the relative standard deviation (rSD) of the triplicate CPM values was verified to be less than 30%. The Hill equation was used to convert CPM values into CTX concentrations.

Data analysis was performed in GraphPad Prism version 6.0 (San Diego, CA, USA), allowing the determination of the parameters of each CTX3C competition sigmoidal curve (using 4 parameters). The detection limit (LOD) and the limit of quantification (LOQ) of the CTX-like composite in fish muscle and liver samples were determined based on the relative standard deviation (rSD) of the top plateau (Bmax) of the dose-response curve according the method described by the authors of [64], following the equations: LOD = Bmax–3 \* rSD; and LOQ = Bmax–10 \* rSD. The values of LOD and LOQ were not estimated in this study. However, previous publications have reported LOQ values corresponding to 0.32 ng CTX3C equiv. g−<sup>1</sup> [27] and 1.5 ng CTX3C equiv. g−<sup>1</sup> [36] and LOD values of 0.75 ng CTX3C equiv. g−<sup>1</sup> [64].

#### 4.5.2. LC-MS/MS

A calibration curve was calculated from the successive dilutions of a CTX3C standard solution (Wako, Tokyo, Japan) in MeOH at concentrations ranging from 12.5 ng mL−<sup>1</sup> to 200 ng mL−1. The limit of detection (LOD) and limit of quantification (LOQ) were calculated statistically based on the following formulae: LOD = 3.3 \* std/S; and LOQ = 10 \* std/S, where "std" is the standard deviation and "S" the slope of the calibration curve composed of successive dilutions of the CTX3C reference material. The calculated values for LOD and LOQ corresponded to 0.57 ng CTX3C mL−<sup>1</sup> and 1.74 ng CTX3C mL−1, respectively. This was equivalent, respectively, to 0.22–1.14 ng CTX3C mL−1, and 0.67–3.48 ng CTX3C g−<sup>1</sup> of lionfish liver, and to 0.54–1.54 and 1.64–4.71 ng CTX3C g−<sup>1</sup> of parrotfish *Chlorurus microrhinos* flesh.

#### 4.5.3. CBA-N2a

Net absorbance data were used to establish the full sigmoidal dose-response curves for CTX3C, lionfish liver extracts, and parrotfish *Chlorurus microrhinos* toxic food. The CTX3C and fish samples were tested in parallel in the same assay. The maximum concentration of the liver tissue (MCE) that did not induce unspecific mortality on N2a cells was determined in OV<sup>−</sup> and OV+ conditions using the fish control sample from the exposure phase of the experiment. This concentration was further used to establish the concentration range of the fish liver samples to use for the assay in the depuration experiment. The maximum concentration of dry extract (MCE) was defined as 20 mg mL−<sup>1</sup> from the fish liver control at 42 days. The limit of detection (LOD) and the limit of quantification (LOQ) of the CTX-like toxicity in the fish liver samples, expressed in ng CTX3C equiv. g−<sup>1</sup> of fish liver, were determined according to the following equations: LOD = (EC80/MCE) and LOQ = (EC50/MCE), where EC80 and EC50 are the values obtained for CTX3C toxin standard, and were determined at 0.06 ± 0.01 and 0.12 ± 0.02 ng CTX3C equiv. g−1, respectively. The composite cytotoxicity in the fish liver samples, expressed in ng CTX3C equiv. g−<sup>1</sup> of fish liver, was then estimated based on the (EC50 of CTX3C/EC50 of fish liver) equation.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/ 10.3390/toxins13080564/s1, Figure S1: Dose-response curves of N2a cells when exposed to increasing concentrations of parrotfish flesh extracts (toxic food) in OV<sup>−</sup> (open symbols) and OV<sup>+</sup> (solid symbols) conditions at 85.7/8.75 μM (final concentrations). Data represent the mean ± SD of each

aliquot tested, with each point run in triplicate. Absorbance values were measured at 570 nm via the MTT assay, after a 45 min MTT incubation time. The initial cell viability was 1.054 ± 0.020 in the RCV control. The mean final cell viability was 0.927 ± 0.023 in the absence of O/V treatment (COV−), and 1.031 ± 0.029 in the presence of non-destructive O/V treatment (COV+), respectively. The dotted vertical line corresponds to the MCE established at 10,000 mg mL−<sup>1</sup> of fish flesh extracts, avoiding non-specific cy-totoxicity in both conditions of OV treatments. The LOD and LOQ values in fish flesh were 0.03 <sup>±</sup> 0.01 and 0.06 <sup>±</sup> 0.02 ng CTX3C equiv. g−<sup>1</sup> [66], Figure S2. The sigmoidal dose-response curve of r-RBA was used to quantify the concentration of the CTX in the fish liver and muscle during the experiment (GraphPad Software, Inc., La Jolla, CA, USA).

**Author Contributions:** Conceptualization, M.-Y.D.B., L.L.M.J., I.d.P.L. Methodology, M.-Y.D.B., I.d.P.L., K.S., A.T., H.B.G., F.O., J.V., H.T.D., M.C., L.L.M.J. Formal Analysis, I.d.P.L., J.V., H.T.D., M.C., L.L.M.J., M.-Y.D.B. Funding acquisition, M.-Y.D.B., M.C. Project administration, M.-Y.D.B., Supervision, M.-Y.D.B., L.L.M.J. Investigation, M.-Y.D.B., I.d.P.L., K.S., A.T., H.B.G., F.O., P.S., J.V., H.T.D., M.C., L.L.M.J.; Writing—Original Draft Preparation, I.d.P.L.; Writing—Review & Editing, I.d.P.L., M.-Y.D.B., L.L.M.J., J.V., H.T.D., M.C., A.T., K.S., H.B.G., F.O., P.S. All authors have read and agreed to the published version of the manuscript.

**Funding:** The research was funded by the countries of France and French Polynesia in the framework of the CARISTO-Pf" (no. 7937/MSR/REC of 4 December 2015 and Arrêté no. HC/491/DIE/BPT of 30 March 2016) research program. This study was also partially supported by the International Atomic Energy Agency (IAEA) through the Coordinated Research Project (CRPK41014), contract #18827 ("Bentox" Project) and the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES, Brazil) for the Ph.D. scholarship awarded to I.d.P.L. at Federal University of Paraná, and for the PVEX (n. 88881.172853/2018-01) and PDSE-2018 (n. 7338918) grants awarded to L.L.M.J. and I.d.P.L., respectively.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** The data presented in this study are available in this article and Supplementary Materials.

**Acknowledgments:** The IAEA is grateful to the Government of the Principality of Monaco for the support provided to Environment Laboratories. Our sincere gratitude to Philipp Hess for hosting L.L.M.J. in the Laboratoire Phycotoxines (ODE\DYNECO) at IFREMER (Nantes, France) and providing the instruments and materials for the LC-MS/MS analysis.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**

