*Article* **Depuration Kinetics and Growth Dilution of Caribbean Ciguatoxin in the Omnivore** *Lagodon rhomboides:* **Implications for Trophic Transfer and Ciguatera Risk**

**Clayton T. Bennett 1,2 and Alison Robertson 1,2,\***


**Abstract:** Modeling ciguatoxin (CTX) trophic transfer in marine food webs has significant implications for the management of ciguatera poisoning, a circumtropical disease caused by human consumption of CTX-contaminated seafood. Current models associated with CP risk rely on modeling abundance/presence of CTX-producing epi-benthic dinoflagellates, e.g., *Gambierdiscus* spp., and are based on studies showing that toxin production is site specific and occurs in pulses driven by environmental factors. However, food web models are not yet developed and require parameterizing the CTX exposure cascade in fish which has been traditionally approached through top-down assessment of CTX loads in wild-caught fish. The primary goal of this study was to provide critical knowledge on the kinetics of C-CTX-1 bioaccumulation and depuration in the marine omnivore *Lagodon rhomboides*. We performed a two-phase, 17 week CTX feeding trial in *L. rhomboides* where fish were given either a formulated C-CTX-1 (*n* = 40) or control feed (*n* = 37) for 20 days, and then switched to a non-toxic diet for up to 14 weeks. Fish were randomly sampled through time with whole muscle, liver, and other pooled viscera dissected for toxin analysis by a sodium channel-dependent MTT-based mouse neuroblastoma (N2a) assay. The CTX levels measured in all tissues increased with time during the exposure period (days 1 to 20), but a decrease in CTX-specific toxicity with depuration time only occurred in viscera extracts. By the end of the depuration, muscle, liver, and viscera samples had mean toxin concentrations of 189%, 128%, and 42%, respectively, compared to fish sampled at the start of the depuration phase. However, a one-compartment model analysis of combined tissues showed total concentration declined to 56%, resulting in an approximate half-life of 97 d (R<sup>2</sup> = 0.43). Further, applying growth dilution correction models to the overall concentration found that growth was a major factor reducing C-CTX concentrations, and that the body burden was largely unchanged, causing pseudo-elimination and a half-life of 143–148 days (R2 = 0.36). These data have important implications for food web CTX models and management of ciguatera poisoning in endemic regions where the frequency of environmental algal toxin pulses may be greater than the growth-corrected half-life of C-CTX in intermediate-trophic-level fish with high site fidelity.

**Keywords:** *Lagodon rhomboides*; pinfish; bioaccumulation; depuration; ciguatoxin; Caribbean ciguatoxin; ciguatera; growth dilution; model; kinetics

**Key Contribution:** This study demonstrates that decreases in C-CTX concentration in fish tissues relate to organismal redistribution across tissues and are a factor of growth dilution, rather than systemic elimination. We demonstrate the importance of evaluating growth dilution during longterm studies of CTX and other natural toxins to avoid underestimates of toxin body burden and overestimates of depuration rates, which have clear implications for trophic transfer and ciguatera poisoning risk.

**Citation:** Bennett, C.T.; Robertson, A. Depuration Kinetics and Growth Dilution of Caribbean Ciguatoxin in the Omnivore *Lagodon rhomboides:* Implications for Trophic Transfer and Ciguatera Risk. *Toxins* **2021**, *13*, 774. https://doi.org/10.3390/ toxins13110774

Received: 27 September 2021 Accepted: 29 October 2021 Published: 1 November 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**

Benthic dinoflagellates have the capacity to produce a diverse suite of bioactive secondary metabolites that have been linked with seafood safety and human health concerns globally. One such group includes the neurotoxic ciguatoxins (CTXs) that have been linked to ciguatera poisoning and have been associated with some species and strains of epibenthic dinoflagellates from the genus *Gambierdiscus* and *Fukuyoa* [1–5]. Several studies have reported that the in situ dinoflagellate community assemblage can change with environmental factors (e.g., temperature) and that species and strains may have physiological niches [6–8]. Likewise, long-term field studies have demonstrated that the CTX load of field-collected benthic microalgae is: (1) asynchronous with *Gambierdiscus* abundance; (2) site specific; (3) seasonal; and (4) occurring in pulses in tropical reef ecosystems [6,9]. It has also been proposed that CTX environmental pulses are linked to the presence of highly toxic *Gambierdiscus* strains rather than high overall algal biomass [5,6,9,10] potentially reducing the effectiveness of genus-level monitoring for these benthic HABs in terms of risk reduction. This presents some questions on the subsequent CTX load and CTX pulses that may occur in other demersal marine biota feeding on potentially toxicogenic epiphytic algae, aquatic invertebrates (e.g., amphipods), and other small grazers.

While still a working hypothesis in the field, the bioaccumulated CTX load in fish collected from ciguatera poisoning hotspots (and therefore ciguatera risk) is a function of the rate of toxin production by epi-benthic dinoflagellates (conceptual model in Lewis et al. [11] and revisited in Lewis and Holmes [12]). However, there remains a gap in understanding the timing between these processes. Field-based studies have provided evidence of a temporal lag of months to years occurring between the environmental triggers that increase toxin production, toxicity in upper-trophic-level fish, and increasing ciguatera cases [13,14]. Further, high fish toxicity (and resulting ciguatera prevalence) has been observed in highsite-fidelity fish when toxigenic dinoflagellate abundance and toxicity are low (or absent), suggesting a shift towards a non-toxic area [15,16], which leads to an assumption that bioaccumulated CTX is persistent in fish after the dinoflagellate source declines. However, field studies lack the ability to properly investigate ecologically relevant exposure routes (e.g., dietary, respiration, dermal) or the rates of uptake and depuration once a CTX source is removed, because total source removal can only be assumed.

Early studies attempted to investigate the retention of CTX-like toxicity (as a proxy for CTX, which was yet to be structurally elucidated) in a laboratory setting using fish that had naturally incurred toxin. For example, in Hawaii, Takata and colleagues (reported by [17]) allowed wild-caught Lutjanids (*Lutjanus bohar*, *L. gibbus*) and a Serranid (*Variola louti*) to depurate in aquariums at various intervals up to 14 months, while Banner et al. [17] kept wild *L. bohar* in holding ponds up to 30 months. In both studies, fish at the end remained toxic when fed to cats and mongoose. Davin et al. [18] fed several species of marine and freshwater fish the ground flesh or extracts of barracuda (*Sphyraena*) or whole *Gambierdiscus* cells (both sourced near the Caribbean Antilles Islands), followed by depuration time up to 81 d. Many of the fish died or were intoxicated. Largemouth bass (*Micropterus salmoides*) fed a high dose of cells recovered behaviorally but remained moderately toxic by intraperitoneal injection to mice after 81 days of clean feed, but fish fed lower doses were non-toxic after depuration. Recently, more laboratory studies have focused on CTX elimination using several exposure methods and in a variety of organisms. For instance, Ledreux et al. [19] performed single-oral dose-recovery experiments with mullet (*Mugil cephalus*) fed *G. polynesiensis* cells and reported only 5% of the CTX activity occurred in tissues after 24 h when analyzed by the mouse neuroblastoma assay (N2a). Li et al. [20] described multiple tissue kinetics of three Pacific CTX congeners, namely P-CTX-1 (CTX1B), P-CTX-2 (52-epi-54-deoxyCTX1B), and P-CTX-3 (54 deoxyCTX1B) extracted from eel (*Lycodontis javanicus*). Toxins were added to a pelleted feed and given to juvenile orange-spotted grouper (*Epinephilus coioides*) for 30 d and subsequently depurated for 30 d. The authors reported that CTX declined exponentially in depuration in some tissues (including muscle); however, CTX burden by the end of the study was not significantly different from

the levels measured prior to depuration [20]. In another recent report, Caribbean CTX (C-CTX-1) depuration was investigated in the muscle of the freshwater goldfish (*Carassius auratus*) following a 43 d daily feeding of naturally C-CTX incurred amberjack (*Seriola* sp.) flesh prepared in agarose [21]. Estimated CTX toxicity (concentration) in muscle was reported to decline by approximately 86% out to 60 d post-exposure, but other tissues were not analyzed. Most recently, a depuration experiment using juvenile lionfish (*Pterois volitans*) was reported after 30–41 d experimental feeding on the flesh of the naturally P-CTX-contaminated parrotfish (*Chlorurus microrhinos*) [22]. Extracts from pooled liver showed a gradual decline in CTX concentration in fish harvested during a 43 d depuration, where fish were switched to a non-toxic diet of farmed sea bream (*Sparus auratus*). These two latter studies report depuration in single tissues (muscle and liver, respectively) and provide informative data on trends in bioaccumulation and depuration following oral exposure but were unable to capture cross-tissue distribution and did not examine fish growth that may contribute to the change in CTX tissue concentrations, but not body burden, as suggested by Holmes et al. [23] in a recent field-based study and review. We propose that estimates of toxicokinetic rates that also incorporate growth dilution may be more valuable than CTX concentrations in tissues in studies aiming to predict or model CP risk in the food web across species.

In this study, we provide data on the kinetic rate of C-CTX-1 bioaccumulation and depuration in the ecologically relevant marine omnivore *Lagodon rhomboides* (pinfish). Replicate fish were either fed a C-CTX-1 (*n* = 40) formulated diet or matrix matched control (*n* = 37) for 20 d, and then placed into depuration for up to 99 d. CTX toxicity in the major tissue compartments of *L. rhomboides* were quantified and compared through time and across replicates compared to control fish using an in vitro mouse neuroblastoma (N2a-MTT) assay.

Sampling and CTX analysis of whole muscle, liver, and other pooled visceral contents (heart, spleen, gall bladder, intestine) throughout the time course revealed dynamic trends in tissue burden and kinetics. Further, modeling concentrations corrected for fish growth dilution revealed an increase in the estimated half-life of C-CTX-1 in *L. rhomboides*, highlighting growth as a major source of pseudo-elimination of accumulated CTX. This correction reveals that C-CTX (measured here as a CTX3C equivalents) can be retained for several months following removal from the toxin source and if not accounted for could result in error during CP risk assessment and analysis of food web transfer potential. This work should be considered when sampling mid-trophic-level fish with high site fidelity and in the development of temporal models of CTX cascades where the frequency of toxin pulses may be shorter than the CTX half-life in fish. These data have implications for CP risk analysis in fish and may help to explain high toxin loads in highly migratory species collected in non-endemic regions for ciguatera (e.g., barracuda, mackerel, amberjack).

#### **2. Results**

#### *2.1. Experimental Design, Diet Formulation and Consistency*

Two dual-phase experimental trials were conducted to assess the depuration of Caribbean CTX-1 (C-CTX-1) that had bioaccumulated in *Lagodon rhomboides* fed a control or low-dose CTX diet through time and were sampled according to Table 1. A laboratory formulated diet was prepared and analyzed for procedural consistency (Table 2). Parameters including water content, pellet weight, whole-food CTX concentration, and pellet CTX concentration were quantified in all three batches of CTX pellets and four batches of control food for each trial. Slight differences were recorded in the water content of pellets after drying; however, food preparation was identical and resulted in the equivalent pellet weights. See methods for full details on the preparation of the control and CTX formulated diet.


**Table 1.** Number of replicates (*n*) Control and ciguatoxin (CTX)-exposed *Lagodon rhomboides* sampled during two trials where CTX-exposed individuals were fed pelleted Caribbean CTX-1 at 0.02 ng CTX3C eq. g−<sup>1</sup> fish (initial weight) day<sup>−</sup>1. Days in parentheses are number of days in the depuration

**Table 2.** Quality control of experimental treatments using the formulated pellet diet. Values are provided as the mean ± standard deviation (s.d.) of subsamples from each batch (per trial, control = 4, CTX = 3).


<sup>a</sup> Calculated from the weight of dry ingredients in the whole food after drying. <sup>b</sup> Calculated from the mean weight of 40 pellet subsamples per batch. <sup>c</sup> Calculated using the CTX3C eq. concentration of dried toxic barracuda powder and amount of powder in the mixture. <sup>d</sup> Calculated from the average weight of pellets provided daily and weight of fish on day 0. <sup>e</sup> Calculated using pellet water content and percentage of initial weight fed daily (whole). <sup>f</sup> Calculated from total weight of pellets provided daily, pellet CTX3C eq. concentration, and initial weight of fish.

#### *2.2. Fish Growth*

The mean initial weight of fish in the first trial (24.4 ± 9.4 g) was smaller than the second (36.4 ± 5.9 g; *p* < 0.0001); however, feeding rate was normalized to each fish at the start of trials so that fish consumed proportionally similar food and CTX relative to their mass (Table 2). At initial, fish were fed approximately 1.8% by weight equaling an exposure rate of around 0.018 ng CTX3C eq. g−<sup>1</sup> fish day−<sup>1</sup> that declined to 0.015 ng CTX3C eq. g−<sup>1</sup> fish day−<sup>1</sup> due to fish growth by day 20 when CTX feeding was stopped. The growth rate constants (*kg*) of control and CTX-treated fish collected at the same time point were similar except on experimental day 20 in both trials (see Supplementary Table S1). This effect was resolved when CTX-fed fish were swapped to the non-toxic diet for depuration.

Raw *kg* determined for the combined experiment between treatments were nonparametric according to statistical comparisons and visual inspection of the residual plots highlighted that the variability on day 6 and 10 d were responsible. Growth rates between replicate fish stabilized from day 10 to 20 (bioaccumulation) as fish stabilized to their control and experimental feeding regime. Exclusion of the data points in the early portion of the trial (day 1–10) restored normality and homoscedasticity. An ANOVA and post hoc analysis showed that the *kg* on days 20 and 25 were significantly different than days 90 and

119 in the CTX fish (Supplementary Figure S2B). The *kg* at day 40 was a transition period and was not significantly different than other time points. Non-linear regression using a one-phase exponential decay model to the *kg* of fish against time (R<sup>2</sup> = 0.70) showed *kg* neared a plateau around experimental day 55 at approximately 4.07 × <sup>10</sup>−<sup>3</sup> day−<sup>1</sup> (see Figure 1). The mean weight of CTX and control fish on the last day were 78 ± 13% and 65 ± 17% larger than at the beginning of the study which caused the daily feeding rate to decrease from around 1.8% to 1.0% of fish biomass by the end of the experiment.

**Figure 1.** Non-linear regression of growth rate (*kg*) and time course in days of CTX fish in depuration.

#### *2.3. Toxin Distribution in the Bioaccumulation Phase*

Maximum concentrations of extracts used for CTX quantification were 50, 10, and 2.5 mg tissue equivalents (TE) well−<sup>1</sup> (217.4, 43.5, and 10.9 mg TE mL−1) for muscle, liver, and viscera samples, respectively, and the mean value of the 75% maximal effective concentration (EC75) from CTX3C dose–response curves was 0.452 ± 0.151 pg well−<sup>1</sup> (1.965 ± 0.657 pg mL<sup>−</sup>1). Based on these data the limit of quantification (LOQ) for each tissue was 0.009 ± 0.003 (muscle), 0.045 ± 0.015 (liver), and 0.181 ± 0.060 (viscera) ng CTX3C eq. g−<sup>1</sup> tissue (Supplementary Figure S1). Brain was extracted and analyzed on a screening basis up to a 30 mg TE dose (130.4 mg TE mL−1) on the N2a-MTT, but all were non-toxic. Likewise, only 9 fish had gonads that were mature while all others were immature and unrecognizable. Mature gonads were extracted and assayed up to a 10 mg TE (43.5 mg TE mL<sup>−</sup>1) and were also non-toxic, therefore brain and gonads were excluded from further study. Quantifiable CTX3C eq. concentrations were possible in viscera across all sampling points, whereas the CTX toxicities in other tissues were detectable but not quantifiable (based on our QAQC criteria) until later time points.

For instance, CTX was detected on the first sampled time point (day 6) in all muscle extracts, but only one fish reached levels exceeding the LOQ. Likewise, CTX was not detected in 3 of the 4 fish livers from day 6 due to non-specific matrix effects on the N2a-MTT but were quantifiable from one replicate fish. After 10 d of CTX meals, 2 of 4 fish had CTX detectable in muscle and 2 of 4 fish sampled could be quantified. After 20 d of CTX feeding, all tissue extracts of sampled fish had toxicity above the determined LOQ and were the highest measured during the CTX feeding phase.

Mean CTX3C eq. concentrations in all tissues of fish harvested on day 20 were not different between trials based on a *t*-test (Table 3). Mean CTX3C eq. concentrations were significantly different in muscle on day 20 of trial 1 and trial 2 compared to day 6 (*p* < 0.05 and *p* < 0.001, respectively) and on day 20 (trial 2) compared to day 10 (*p* < 0.01; α = 0.05; Supplementary Figure S3A), but no statistical difference was found in liver or viscera of fish sampled during the bioaccumulation phase (day 6–20) based on ANOVA (Supplementary Figure S3B,C). A steady state of CTX concentration was not reached by day 20 based on the continued upward trend in CTX through the feeding period (Figure 2). Since this was not the goal of the present study, this was an acceptable outcome.

**Table 3.** Between-trial reproducibility of CTX-1 concentrations measured as CTX3C eq. by mouse neuroblastoma assay (N2a-MTT) in fish tissues collected at day 20. Data for each trial are the mean ± s.d. (*n* = 4). All controls were negative and are not shown. A two-tailed *t*-test was performed to test for significant differences (α 0.05; df = 6).


**Figure 2.** Linear regression of CTX-1 measured as a CTX3C eq. concentration (mean ± s.d.) in sampled (**A**) muscle, (**B**) liver, and (**C**) viscera through bioaccumulation (red) and depuration (blue). Solid regression lines denote a significant correlation while dashed lines were not significant due to high variability between individual fish within a tissue type. Control fish were negative for CTX in all cases and are not plotted. Data points represent the mean ± s.d. of replicate fish (*n* = 4 except on day 20 where *n* = 8 across two trials).

#### *2.4. Toxin Distribution in the Depuration Phase*

The primary goal of this study was to estimate depuration and based on the consistency of CTX3C eq. concentrations in fish analyzed at day 20 (see Table 3), we had confidence in combining sample data from the two identical trials for an extended depuration phase analysis (details in methods; summary in Table 1).

Mean CTX3C eq. concentration in muscle extracts were not significantly different between 0 and 70 d into the depuration period (Supplementary Figure S3A). The highest mean muscle CTX concentrations were measured in fish 99 d into depuration (day 119 of the complete study) and was 0.09 ng CTX3C eq. g−<sup>1</sup> TE (Figure 2A). The mean concentration in muscle on the last day of depuration was significantly higher than those measured at all prior time points based on ANOVA, except for the day 20 fish from trial 2 which only marginally fell outside the criteria for significance (α = 0.05; *p* = 0.054; Supplementary Figure S3A). Consistent with these results were lower extract doses on the N2a-MTT needed for fish sampled on day 119 (5–20 mg TE) versus fish sampled earlier in the depuration phase (30–50 mg TE).

Liver extract CTX3C eq. concentrations from replicate fish in the depuration phase were the most variable of the tissues analyzed with relative standard deviation for trials 1 and 2 between 60 and 100% and 52 and 99%, respectively. This variability was consistent during both trials and through each sampling point with no significant trends (R2 = 0.09; *p* = 0.152; Figure 2B). Concentrations in liver of fish from trials 1 and 2 during depuration ranged from 0.12 to 1.77 and 0.23 to 2.52 ng CTX3C eq. g−<sup>1</sup> TE, respectively, and no statistical differences were found between days 20 and 119 (α = 0.05; Supplementary Figure S3B).

Mean toxicity of viscera extracts were highest in fish sampled 5 d into the depuration phase (5.61 ± 1.06 ng CTX3C eq. g−<sup>1</sup> TE; Figure 2C). No significant loss of CTX from viscera was measured by the end of trial 1 (day 20 = 4.40 ± 1.17 vs. day 40 = 4.43 ± 1.34 ng CTX3C eq. g−<sup>1</sup> TE; Supplementary Figure S3C). When depuration time was extended in trial 2, a significantly lower concentration was measured on day 119 (1.73 ± 0.21 ng CTX3C eq. g−<sup>1</sup> TE) than was measured 5 d and 10 d into depuration (day 25 vs. 119 *p* < 0.01; day 30 vs. 119 *p* = 0.018; Supplementary Figure S3C).

#### *2.5. Muscle, Liver, and Viscera CTX Kinetics*

Linear regression of the bioaccumulation phase data showed a positive correlation between the number of exposure days and CTX3C eq. concentration for muscle and viscera (R<sup>2</sup> = 0.70, *p* = 0.0001; R<sup>2</sup> = 0.47, *p* < 0.01, respectively; Figure 2A, and 2C; Table 4). The best fit models for bioaccumulation in both cases were simple straight-line equations (muscle and viscera AICc = 100 and 99.4%, respectively) in the form of y = *kuptake* ∗ x + b, where *kuptake* is uptake rate (ng CTX3C eq. g−<sup>1</sup> fish day−1) equal to 3.37 × <sup>10</sup>−<sup>3</sup> for muscle and 0.124 for viscera (Table 4). Some liver samples on days 6 (*n* = 3) and 10 (*n* = 2) were excluded from the regression analyses due to significant non-specific matrix issues on the N2a-MTT assay because a value of zero could not be assumed. Liver concentrations during accumulation followed a non-significant increase (R<sup>2</sup> = 0.24, *p* = 0.129; Table 4), so an uptake rate is not reported.

**Table 4.** Results of linear regression analysis for CTX3C eq. concentrations in fish samples during the experimental bioaccumulation and depuration phases.


Straight line and one-phase exponential decay models were chosen for comparison of depuration kinetics of individual tissues. Depuration resulted in a linear decrease in viscera CTX3C eq. concentration (R<sup>2</sup> = 0.44; *p* = 0.0005; Figure 2C), but an unexpected linear increase occurred with depuration time for muscle (R2 = 0.43; *p* = 0.0005; Figure 2A). The regression was likely skewed because of the weight of day 119 on the end of the regression line. Model comparison to describe depuration kinetics of tissues also showed simple linear models were the most probable in each case and depuration rates for muscle and viscera were −4.62 × <sup>10</sup>−<sup>4</sup> and 3.31 × <sup>10</sup>−2, respectively (Table 4). However, liver CTX concentrations did not follow any pattern of depuration (R<sup>2</sup> = 0.09; *p* = 0.152), so a depuration rate is not reported here.

#### *2.6. Muscle, Liver, and Viscera CTX Burdens*

Tissue burden, i.e., the amount of contaminant per tissue (ng CTX tissue−1) was calculated by multiplying the CTX concentration in each tissue by the whole tissue wet weight determined at the time of dissection. The concentrations of CTX (ng CTX g−<sup>1</sup> TE) within *L. rhomboides* tissues were present at levels of descending concentration as follows: viscera > liver > muscle. However, the tissue CTX burden (that incorporates total tissue mass) showed some variability between individuals. For instance, viscera carried the highest toxin burden in all fish fed CTX throughout the entire experiment followed by muscle in 28 individuals. In the other 12 CTX-exposed individuals, liver CTX burden was slightly higher (1.4 ± 0.3 fold) than in muscle for 10 fish. CTX was not detected in muscle and liver of two fish at day 6 which prevented a comparison.

Total CTX body burdens in fish were not directly compared across days because fish were fed different amounts of CTX based on individual starting body weight. Instead, toxin burdens of replicate fish were compared by the relative distribution of CTX as a percent of the sum of toxin burden measurements in tissue compartments (Figure 3). Up to day 40, viscera contained most of the total measured toxicity (89 ± 9%) followed by muscle (7 ± 4%) and liver (3 ± 2%). From day 40 to 119, the relative distribution of toxicity began to shift. While relative viscera burden declined to 44% the relative muscle burden increased to 41% on the last day.

**Figure 3.** Distribution of the C-CTX-1 burden in tissues of *L. rhomboides* expressed as a percent of the whole burden measured in CTX3C eq. Note that fish were placed into depuration after day 20. Bars represent the mean ± s.d. of replicate fish (*n* = 4 except on day 20 where *n* = 8 across two trials). \* CTX was below detectable levels on day 0.

#### *2.7. One-Compartment Model Kinetics and Growth Correction*

For the one-compartment model, the total concentration (*Cfish*) represents the sum of CTX burdens in muscle, liver, and viscera divided by the combined mass of those tissues. The one-compartment model showed CTX bioaccumulation in *L. rhomboides* followed a linear increase from day 0 to 20 at an overall *kuptake* of 8.80 × <sup>10</sup>−<sup>3</sup> ng g−<sup>1</sup> day−<sup>1</sup> (R2 = 0.67, *p* = 0.0001; Figure 4).

**Figure 4.** One compartment analysis of accumulated C-CTX-1 quantified by N2a-MTT as a PCTX3C eq. showed bioaccumulation (red) followed a linear pattern, while depuration (blue) followed exponential decay. The uptake rate (*kuptake*) calculated by linear regression was equal to 8.80 <sup>×</sup> <sup>10</sup>−<sup>3</sup> ng CTX3C eq. g−<sup>1</sup> day−<sup>1</sup> (R2 = 0.67, *p* = 0.0001) and logarithmic transformation of depuration data resulted in a first-order relationship (R2 = 0.43). The overall elimination rate constant (*k*2) was 7.161 <sup>×</sup> <sup>10</sup>−<sup>3</sup> day−<sup>1</sup> based on the slope of the linear plot. Vertical dashed lines indicate the point where fish were swapped from CTX feed to a non-toxic food for the depuration phase. Data points represent the mean ± s.d. of replicate fish (*n* = 4 except on day 20 where *n* = 8 across two trials).

The straight-line plot of the transformed total concentration data (Ln [*Cfish*]) against time confirmed depuration approximately followed first-order kinetics (R<sup>2</sup> = 0.43) and the overall *<sup>k</sup>*<sup>2</sup> (linear slope) was equal to 7.161 × <sup>10</sup>−<sup>3</sup> with an estimated half-life of 97 d. Significant fish growth that was measured during the depuration phase meant the kinetic rate of elimination needed to be normalized to the CTX dilution rate due to increasing fish mass, a source of pseudo-elimination. Growth correction of *k*<sup>2</sup> was investigated through multiple exponential decay simulations of the depuration phase data, where measured *k*<sup>2</sup> is adjusted using measured fish growth rates (*k*<sup>2</sup> <sup>−</sup> *kg* = *k*<sup>2</sup> *growth-corrected*) and based on estimates of initial CTX3C eq. concentrations (*Cfish(i)*) immediately prior to depuration. The combined mass of whole muscle, liver, and viscera tissues (∑ [muscle + liver + viscera] g) had a strong linear correlation with whole body mass (R<sup>2</sup> = 0.95; Figure 5) during the study, as did total accumulated burdens with cumulative doses at the end of the bioaccumulation phase (day 20) (R<sup>2</sup> = 0.95; Figure 6) and provided a good estimate of initial CTX3C eq. concentrations (estimated *Cfish(i)* = total CTX3C eq. burden/sum mass of whole muscle, liver, and viscera) in each fish that entered depuration.

**Figure 5.** Combined sum of measured tissue masses (∑ [muscle + liver + viscera]) (black) of all *L. rhomboides* sampled in this study as a function of total fish mass. Total body masses at the start of depuration were estimated using an exponential growth formula (see Methods) and were used to estimate the ∑ [muscle + liver + viscera] (red) for each fish at experiment day 20 which entered the depuration phase (*n* = 24). Dashed lines are 95% confidence intervals.

**Figure 6.** Linear correlation of cumulative CTX burden with total dose in *L. rhomboides* sampled during the bioaccumulation phase (black; *n* = 16; R<sup>2</sup> = 0.95). The relationship was used to estimate the CTX burden at the start of the depuration phase (i.e., after day 20) for fish remaining after the bioaccumulation period that entered depuration (red; *n* = 24). Dashed lines are 95% confidence intervals. Error bars are standard deviation of the CTX burdens (measured concentration ± s.d. × tissue mass). Some points overlap.

In the Simulation Model (Figure 7A), regression analysis with logarithmic transformation of the predicted *Cfish(t)* values for each depuration time point (Ln [*Cfish(t)*] *= -k2* ∗ *t +* Ln [*Cfish(i)*]) resulted in a predicted half-life of 75 d (R2 = 0.70; compared to 97 d from measured *Cfish(t)*) and a correlation showed the Simulation Model fit closely to the measured *Cfish(t)* (R<sup>2</sup> = 0.68; *p* = 0.02). The Simulation Model was used as a starting point whereby several models were compared to assess the effect of growth. In Model 1, growth correction of the Simulation Model using the replicate average *kg* at each sampled time point (Figure 1; Table 5) increased the half-life to approximately 143 d based on model fitting parameters (Model 1: *<sup>k</sup>*<sup>2</sup> *growth-corrected* = 4.84 × <sup>10</sup>−3; R2 = 0.364; Figure 7B). Growth correction for Model 2 was performed using three growth rates based on the ANOVA and Tukey's test (Table 6; Supplementary Figure S1B) and produced a similar half-life at 148 d (Model 2: *<sup>k</sup>*<sup>2</sup> *growth-corrected* = 4.68 × <sup>10</sup><sup>−</sup>3; R2 = 0.345; Figure 7C). Likewise, growth correction using the *kg* measured for each individual fish resulted in a half-life of approximately 143 d (Model 3: *<sup>k</sup>*<sup>2</sup> *growth-corrected* = 4.83 × <sup>10</sup><sup>−</sup>3; R2 = 0.360; Figure 7D). Lastly, a common approach to growth correction was performed by multiplying the measured concentrations by a correction factor (1 + *kg* ∗ time). The correction factor method resulted in a half-life of approximately 157 d, although the uncertainty was greater in this model shown by the lower coefficient of determination (correction factor method: *<sup>k</sup>*<sup>2</sup> *growth-corrected* = 4.41 × <sup>10</sup><sup>−</sup>3; R2 = 0.22; Figure 7).

**Figure 7.** Modeled C-CTX-1 concentrations (based on CTX3C equiv.) compared to measured and growth-dilution-corrected models. Complete details on the (**A**) simulation model; (**B**) Model 1; (**C**) Model 2; (**D**) Model 3; are provided in the methods. (**E**) In the correction factor method, measured concentrations were adjusted using a simplistic approach where measured data were multiplied by a growth correction factor (i.e., *Cfish* ∗ (1 + *kg* ∗ *t*<sup>Δ</sup> *total*)). Plotted symbols represent the mean ± s.d. (*n* = 4 except on day 20 where *n* = 8).


**Table 5.** Growth-corrected depuration rate constants (*k*<sup>2</sup> *growth-corrected*) that were used in growthdilution correction Model 1 of this study. The mean *kg* (*n* = 4) of replicate fish at each time point were used for the correction in this model.

<sup>a</sup> Calculated by subtracting average *kg* from the overall *<sup>k</sup>*<sup>2</sup> value of 7.16 × <sup>10</sup>−3.

**Table 6.** The *k*<sup>2</sup> *growth-corrected* that were used in the growth-dilution correction Model 2 of this study. Time intervals where significant differences were detected by ANOVA and a Tukey's test (See Supplementary Figure S2B) were used for correction in this model.


<sup>a</sup> Calculated by subtracting average *kg* from the overall *<sup>k</sup>*<sup>2</sup> value of 7.161 × <sup>10</sup>−3.

#### **3. Discussion**

This study is the first multi-tissue report on the dynamics of C-CTX elimination in an ecologically relevant intermediate consumer from the greater Caribbean region. This also represents the longest experimental C-CTX fish depuration in an important and rapidly expanding area of research. Since the initiation of these experiments in 2019, at least three similar experiments have been undertaken [20–22] and recently a conceptual model based on field-collected fish, focused on the contribution of growth in lowering CTXs was released [23]. Clearly, the knowledge gaps on CTX bioaccumulation and depuration kinetics have been acknowledged, and several groups are working to fill these. Modeling the disposition of CTX in wild reef fish is a next step for the development of predictive models on the time course of environmental changes, dinoflagellate proliferation and/ or toxification, and CP outbreaks.

#### *3.1. Experimental Considerations and Outcomes*

Pinfish (*L. rhomboides*) are omnivorous and undergo transitional feeding behavior during their life cycle. For instance, juvenile pinfish have a wide-ranging diet (small crustaceans, fish, polychaetes, tunicates, hydroids, seagrasses, epiphytes), while adults are reported to feed predominantly on epiphytes and macrophytes but continue to supplement their diet with many small invertebrates throughout their life [24–27]. As such, a generic trophic-level analysis can be difficult to determine based on the wide dietary range and multiple ontogenetic transitions during the lifecycle [25]. Pinfish are found from the coasts of New England to Brazil, and are widely distributed throughout the Gulf of Mexico, Florida Keys, and Yucatan [28], where CP also occurs. Pinfish also represent a substantial portion of the biomass in seagrass meadows and other structurally complex habitats (e.g., mangrove propagules, reefs, docks) and are a major source of organic matter exported to reef fish during seasonal migrations towards offshore reefs (20–100 m deep) in the Florida Big Bend area [29,30]. The stomach contents of gag grouper, a commercially important species from offshore reefs, contained 47% pinfish during migration periods [31], highlighting this species as a key prey item for higher-trophic-level fishes that have been implicated in ciguatera. Pinfish have also been used to study brevetoxin accumulation and

exposure of many other natural and human contaminants [32]. Wild-collected fish used in this study were collected from an area with low prevalence of CTX and in the absence of brevetoxin-producing *K. brevis* blooms; however, to ensure no naturally incurred levels of CTX or brevetoxin that may skew our results, fish were acclimated, and subsamples of the population were tested for CTX and brevetoxins immediately following collection and the beginning of our experimental trials to ensure no detectable levels.

Some of the differences in our experimental design and the design of prior studies were important for the study goals. There is a trade-off to consider for experiments where it is necessary to maintain equal treatment conditions including the exposure rate (feeding) across large groups of experimental replicates. Since space was limited, alternatives such as grouping animals together in exposure tanks or pooling samples for analytical purposes, were considered, but risked the loss of statistical power and would introduce pseudo-replication. One of our goals were to capture the variability of CTX uptake and depuration in this species, so while fish had some minor size variability at the start of the trials, feed rations based on individual fish mass were prepared and maintained to ensure a consistent daily intake. Now that we have characterized the variability in these fish, we may be able to reconsider these alternatives in future studies to include changes in CTX metabolite formation and distribution via LC-MS/MS strategies (that require greater biomass for the CTX doses used here). The commercial omnivorous fish food that was mixed to create the pellets provided supplemental nutrition and acted as a binding agent which prevented fragmentation and loss of the powderized barracuda during feeding. In other studies, control and treated diets were either not matrix matched [33] or controls were not used [19]. In this case we were able to track similar growth rates between control and treated fish across trials since both of these aspects were controlled. The regulated feeding rate also led to a strong correlation (day 0–20; R<sup>2</sup> = 0.95) between cumulative burdens and the ingested dose in fish sampled during the CTX feeding phase as reported by others [33]. This correlation was used to estimate the accumulated CTX burden in fish prior to their depuration and allowed the analysis of the change in total body concentration of individuals over time under multiple projections using growth correction which was also attempted in one prior experimental study [20]. The isolation of fish, controlled feeding rate, and sufficient replication also are important to capture the individual physiological variability in CTX toxicokinetics, but these have not always been met in other CTX studies [21]. Parallel controls at each sampling time point (except experimental day 90) provided both growth and behavioral assessment (no behavioral issues observed) and analytical controls for the tissue analyses by the N2a-MTT assay.

The procedural consistency across trials was crucial for us to consider CTX distribution data through time as a single set in the analyses. Several points of data support our consistency in approach. Quality control data shows reproducible pellet toxicity and feeding rates across both trials which was supported by the cumulative burden measurements in fish related to the total dose (Figure 6). Further, four fish were analyzed in both trials on the last day of the bioaccumulation phase (day 20) and the toxicity of all tissues sampled were not significantly different (see Table 3). The controlled formulation of the experimental pellet diet also showed that pellet weight and proportional amount of food provided to each fish between control and treated groups was consistent. Based on the supporting evidence collected, we deemed it acceptable to combine datasets from the two experimental trials for analyses of the extended depuration kinetics.

Use of the N2a-MTT assay to determine tissue CTX concentrations has been widely accepted by the CP research community and seafood management agencies alike for several decades [34]. This methodology has been used in prior CTX exposure studies of various organisms to analyze CTX activity over time [19,21,35,36] and provides a measure of the composite toxicity of sodium channel activating metabolites that have implications for human health. In selecting N2a-MTT for our measurements we gained the sensitivity that we needed to evaluate toxicokinetics across individuals through time in this small species but lost the ability to evaluate trends in CTX profiles throughout the experiment. Unfortunately, we had insufficient tissue to support LC-MS/MS analyses without combining replicates (as in other studies) and we wanted to maintain and evaluate the variability between fish so that trends in depuration among tissues would be more robust. As LC-MS/MS methods for CTXs become more sensitive (or with the use of a larger species) we hope to revisit the change in toxin profiles (if any) through these processes in future work.

#### *3.2. Muscle, Liver, and Viscera CTX Kinetics: Bioaccumulation*

Prior to the study, fish were acclimated to the feeding schedule using cooked shrimp and transitioned to gel pellets (no barracuda added) over the 2-week period. By the start of experiments fish were consuming pellets immediately when added to the tanks and continued to do so throughout the entire study. Rapid consumption during the experiment likely limited leaching of toxin from the CTX-pellets into the aquarium water. Additionally, 35% water changes were performed weekly on all aquarium systems to maintain water quality and prevent accumulation of possible residual toxins in the tank water. Although tank water was not analyzed for CTX, the tanks of control and CTX-dosed fish shared the same recirculating artificial seawater by randomized design and none of the control fish tissue extracts or control feed exhibited CTX activity by the N2a-MTT. Additionally, tank substrate was regularly vacuumed by siphoning to prevent the re-uptake of any CTX in fish excrement. Therefore, we assumed with confidence that measured CTX activity was dominated via the dietary exposure route.

During the uptake phase (day 1–20), measured CTX concentrations in muscle and viscera of pinfish were positively correlated with number of CTX exposure days and followed linear trends (Figure 2). Toxin uptake rate estimated in pinfish muscle (0.003 ng CTX3C eq. g−<sup>1</sup> TE day<sup>−</sup>1) by N2a-MTT was in the range of uptake rates reported for P-CTX-2 and -3 in juvenile grouper (*Epinephelus coioides*) measured by LC-MS/MS (0.001–0.005 ng g−<sup>1</sup> day<sup>−</sup>1) but was 10-fold lower than P-CTX-1 (0.033 ng g−<sup>1</sup> day<sup>−</sup>1) [20]. Toxin uptake rate in pinfish viscera (0.123 ng CTX3C eq. g−<sup>1</sup> TE day−1) was not comparable to other studies due to sample preparation of this compartment; however, CTX bioaccumulation in viscera was >41 times the rate in pinfish muscle which led to extremely high visceral concentrations (4.40 ng CTX3C eq. g−1) versus muscle (0.04 ng CT3C eq. g−1) by the end of the bioaccumulation phase.

Between days 6 and 20, CTX concentrations in viscera were 42–330 fold (except 1 fish on day 6) and in liver, 3–69 fold higher than the concentration in muscle. The range of concentration ratios for liver to muscle are similar to what has been reported from fish collected in CP endemic waters [37]. However, as noted by Vernoux et al. [37], we did not detect any pattern in the concentration ratios (data not shown), and therefore support their recommendation to avoid using liver or viscera results to extrapolate flesh toxicity. These compartments may confirm CTX exposure in fish but would not reliably estimate risk of consuming fish fillets.

In some cases, the variability observed in CTX tissue concentration between individual fish replicates in this study was high. This variability was important to capture if we are to extrapolate future field studies where intraspecies variability will be even greater. The strength of the correlation between cumulative CTX burden and total ingested dose supports that fish were provided a normalized amount of CTX and nutritional food, as does our quality assurance/quality control data (Tables 2 and 3). Our time until detection in the muscle and liver using the N2a-MTT assay was much longer than found for mullet, which occurred 3 hrs post-exposure to a single dose of *G. polynesiensis* cells [19] which could relate to species differences or the increased toxicity of P-CTXs produced by *G. polynesiensis* compared to C-CTX-1 that was used in this exposure study. Additionally, our CTX dose was approximately 15-fold lower (0.02 vs. 0.3 ng CTX3C eq. per g of body weight) to ensure that no behavioral disturbances were detected and supports an ecologically relevant range. Intraspecific variability in assimilating and compartmentalizing CTX also could have contributed to the limited detectability of CTXs in the muscle and liver during the first 10 days of the bioaccumulation phase. Omnivorous goldfish fed C-CTX-1 from *Seriola* sp. flesh at a rate of 0.014 ng CTX1B eq g−<sup>1</sup> TE also had undetectable CTX in muscle after the first CTX dose but CTX was measurable (*n* = 2) by day 8 [21]. The dose and physiological differences between herbivorous and omnivorous fish may explain the observed lag time between the CTX activity of benthic dinoflagellates and herbivores compared to uppertrophic-level fish [6,13].

Through this work, we estimated an average net CTX assimilation of 43% in pinfish from the ingested dose throughout the bioaccumulation phase of this study (days 1–20; see Figure 6). The bioavailable amount of the CTX dose, which is the total CTX absorbed into systemic circulation, was not quantified here, but our estimate of net assimilation (bioavailable amount—first pass metabolism elimination) in pinfish was much higher than reported for C-CTX-1 in freshwater goldfish [21], P-CTX-1, -2, and -3 in juvenile grouper [20], and CTX3C in naso [33], which were all less than ~10%. In contrast, mullet assimilated 42% of ciguatoxicity from a single dose of *G. polynesiensis* [19] and orally bioavailable CTX in rats was calculated around 39% [35] which are congruent with measured levels in this study. The CTX burden in mullet declined to 5% of the single dose administered by 24 h, but repeated exposure was not evaluated [19]. It is also possible that C-CTX-1 investigated in this study has a higher assimilation compared to other CTX congeners, among other factors.

Considering the continued linearly increasing trend of CTX in each tissue during the exposure and bioaccumulation phase, we do not assume a steady state was reached during the CTX feeding phase. Clausing et al. [33] determined in herbivorous naso fed *G. polynesiensis* cells, that muscle CTX concentration stabilized somewhere between 8 and 16 weeks of exposure at bioaccumulated levels of approximately 3 ng CTX3C eq. g−<sup>1</sup> TE. However, the CTX concentration in muscle of goldfish stabilized by day 29 at 0.03 ng CTX1B eq. g−<sup>1</sup> TE [21]. This dissimilarity could be due in part to different physiology of freshwater and marine fish, but the possibility that the low sample size (*n* = 2) combined with biological variability prevented detectable trends in the goldfish study cannot be dismissed. Goldfish were reportedly not growing during the study whereas biomass of juvenile naso increased approximately four times compared with initial, therefore, the possibility that there may be differences in the steady-state concentrations reached between growing and non-growing fish cannot be ruled out.

#### *3.3. Muscle, Liver, and Viscera CTX Kinetics: Depuration*

Depuration resulted in an expected decrease in viscera CTX3C eq. concentration with time that followed a simple linear relationship (Figure 2C). Initially, however, the C-CTX-1 (measured as CTX3C eq.) concentrations in viscera continued increasing into the depuration phase and were at the highest measured levels five days after depuration was initiated (day 25 = 5.61 ng CTX3C eq. g−1). A similar trend was reported in a depuration study of juvenile grouper (*Epinephelus coioides*) where P-CTX- 1, -2, -3 concentrations in the skin and intestine were higher 4 days after depuration was started and then followed an exponential decline through the next 30 days [20]. One factor mentioned by the authors was that they observed some undigested material in stomachs when dissected, so this could have had an effect. In this study, we excluded stomachs to avoid possible crossover between assimilated and non-assimilated CTX. We did not measure fecal pellets or the tank water during the trials, so we were unable to capture the CTX levels eliminated from the organism directly; however, controls shared circulating tank water with CTX fish and no controls had detectable CTX levels supporting that dietary uptake was the only relevant exposure route.

Further explanation of the lag of CTX seen in the viscera compartment during this study could be attributed to the amount and quality of protein and the presence of carbohydrates and fiber in the pelleted food. Food quality has been well established as a factor that may increase gut retention and modify absorption kinetics for other toxicants in humans and mammals (see [38–40]), which could be examined in future studies for CTX. Variability of CTX kinetics, particularly from visceral organs, that has been reported between studies may be a function of dietary differences, in addition to species and experimental factors

already described. Persistent viscera contamination could also be influenced by enterohepatic recirculation as seen for anthropogenic contaminants where a xenobiotic is re-absorbed by the intestine after first pass metabolism and returned to systemic circulation [41,42]. Cycling the toxin back through the system would redispose it in tissues, altering toxin compartmentalization dynamics.

CTX depuration in muscle tissue was reported to be rapid in recent studies of juvenile grouper, with half-lives for P-CTX-1, -2 and -3 reported as 28, 26 and 33 days, respectively [20]. Our data on C-CTX-1 depuration in pinfish muscle contrast with these results with C-CTX (as CTX3C eq.) concentrations being significantly higher on the last day than at the start of depuration (Figure 2A), highlighting possible species-level differences. Our data from day 119 likely pull the regression analyses towards a positive slope, but the low variability between replicate fish and significant differences measured between days 20 and 119 showed that an increase in CTX tissue concentration in muscle was real. Further supporting this phenomenon, it should be noted that during long-term studies such as ours, fish are growing, and if CTX levels in muscle were growth-corrected a further increase would be expected. These effects and models were comprehensively explored in this study. Decreasing trends in viscera were paralleled by the increasing CTX levels observed in muscle tissue, suggesting that CTX may be redistributed and/ or compartmentalized into muscle from other tissues through time. A biphasic change in the location of the cyanotoxin cylindrospermopsin was also reported in mussels and attributed to a re-distribution of the toxin within tissues [43]. This may explain the persistent high levels of CTX found in the upper-trophic-level fish in regions with frequent CTX exposure. Analysis of CTX levels in blood between days 90 and 119 may have determined if enterohepatic recirculation of CTX was occurring but was not performed in this study. These data remind us that toxicokinetic models of CTX should consider organ compartments as dynamic places where the toxins have the potential to be in flux and influenced by many environmental and physiological mechanisms, yet to be elucidated. Hopefully, these data will press further studies to investigate the CTX dynamics between tissues, since this may aid in fisheries management and future dockside testing once rapid field assays become available (i.e., knowing which tissues will provide a reliable assessment of CTX risk).

There were no significant trends observed in the liver over time, except for highly variable CTX concentrations throughout the depuration phase. This variability could be attributed to metabolic process regulating elimination from liver. For instance, novel glucuronide metabolites of C-CTX were recently identified in hepatic microsomes indicating that C-CTX glucuronidation may be a prevalent biotransformation pathway in fish [44]. In contrast, grouper efficiently eliminated 90% of the P-CTX from liver by the end 30 days [20]. Additionally, under controlled exposure conditions, pooled lionfish liver samples (*n* = 1 of pooled fish) analyzed by N2a-MTT gradually became less toxic over a 43 d depuration.

#### *3.4. CTX Tissue Distribution*

The relative CTX distribution in *L. rhomboides* tissues allowed us to compare the disposition of CTX burden across fish of different sizes through time (Figure 3). The order of highest to lowest CTX burden during the entire study was viscera > muscle > liver, but prior studies have reported a similar analysis with variable findings. The highest percentage of total CTX was in the muscle of grouper fed 1 ng P-CTX g−<sup>1</sup> fish day−<sup>1</sup> (approximately 50-fold higher dose than the present study) followed by other visceral tissues (intestine and liver combined) for the entire experiment (except day 2 where skin CTX burden was highest) [20]. Our visceral extract contained several tissues not analyzed in the grouper (heart, spleen, gall bladder) which have been shown containing high CTX concentrations in some fish and could explain the differences reported between studies [19]. The CTX burden reported in the muscle of mullet also contained the highest toxin load when fewer than nine CTX meals had been given [19]. However, similar to our findings, data extending beyond nine CTX feedings on dinoflagellate cells by mullet, saw most of the CTX activity detected in the intestine and gall bladder (~50%) followed by muscle (22%) and a small amount in liver (2%) [19]. Likewise, giant clams also contained most of the CTX in the viscera following exposure [36].

#### *3.5. One-Compartment Model Kinetics and Growth Correction*

The C-CTX-1 uptake rate calculated for the combined tissue compartments in *L. rhomboides* was 8.80 × <sup>10</sup>−<sup>3</sup> ng CTX3C eq. g−<sup>1</sup> TE day−1. For comparison, the estimates for P-CTX-1 and -2 uptake rates in whole body of juvenile grouper were 25.2 and 3.51 × <sup>10</sup>−<sup>3</sup> ng g−<sup>1</sup> day−<sup>1</sup> by LC-MS analysis [20]. While our analyses do not account for CTX in some organismal compartments (e.g., kidney, carcass), the included tissues likely contained the large majority of CTX that was bioaccumulated. Toxin accumulation was lower in muscle tissue, which was exhaustively removed from fish during dissection, except for muscle tissue surrounding the head, which was difficult to reliably remove from the small fish and minor in mass. Bioaccumulation in hard parts such as bone and fin are possible but are likely minor contributors to storage and not evaluated in this study. CTX activity was below detection in the brain and gonads so these were excluded from whole body concentrations reported in this study. There is still a lot of room for future studies that will better inform organismal models of CTX bioaccumulation, and these should include longer period of bioaccumulation, different fish species, life stages, and varied CTX congeners, for a more comprehensive analysis of bioaccumulation rates.

Fish growth which contributes to lowered concentration of contaminants, can be a difficult effect to control when investigating depuration, and complicates analyses in exposures when significant [45]. Clausing et al. [33] found that naso growth masked CTX accumulation, causing an apparent steady state in muscle at 8 weeks with similar CTX concentrations occurring at 16 weeks of exposure and significant growth. In that example, growth dilution kept CTX concentrations stable between weeks 8 and 16 while the muscle CTX burden continued to increase. This effect cannot be ignored in the present study which covered a 4 month timeframe. Quantifying the growth-dilution effect has been handled in several ways when studying elimination of contaminants, including adjusting measured concentrations by a growth correction factor as performed by Li et al. [20]. However, these growth adjustment strategies have been critiqued by others and alternative strategies suggested [45,46]. In our study, we applied several growth models and corrections including the one reported by Li et al. [20] to determine the best approach for this and future investigations for C-CTX depuration and growth. We first evaluated the rate constant subtraction method outlined by Brooke and Crookes [45], which eliminates the change in concentration due to growth from the other first-order depuration process (i.e., elimination by respiration, feces, and metabolic transformation). Essential to our rate constant growth correction analyses were the predicted concentrations (*Cfish(i)*) from body CTX burden and weight estimates at the start of depuration. In our case, knowing the feeding rate, dose, and growth rates of all fish (i.e., for *n*= 40 C-CTX treated; and *n*= 37 control) in the study allowed us to perform model analyses that have not been used in past CTX exposure studies. While first order kinetics were assumed based on regression of the transformed concentration data (Figure 4), we acknowledge that a first order relationship is approximate, and is a potential source of uncertainty in the model. However, this method did provide evidence that fish growth was the major contributor to declining CTX concentrations through time in this study.

The growth-corrected half-life for P-CTX-1, -2 and -3 (e.g., 21, 21, and 38 d, and R<sup>2</sup> equal to 0.97, 0.84, and 0.37, respectively) in the whole body of the juvenile grouper *E. coioides* [20] were much shorter than reported by us and others [11,17]. This anomaly could be attributed to unequal toxicokinetics between *E. coioides* and other fish, the use of juvenile specimens, differences between CTX variants, or the statistical power or models applied in that study. Our growth correction modeling technique increased the estimated CTX depuration half-life from approximately 80 d in the raw data simulation to approximately 146 d (Figure 7A–D). Compared to other studies, our growth-corrected half-lives of total ciguatoxicity are more similar to estimates in naturally incurred ciguatoxic Pacific

moray eels (*L. javanicus*) and red snapper (*L. bohar*) where half-lives of 264 and 900 d were reported, respectively [11,17]. Comparison to these data would depend on whether the fish grew significantly, but if growth was positive (as would be expected), growth correction would increase the reported half-life. From this study and others, we might expect a half-life of several months or longer which would be in line with long-term patterns of persistent ciguatoxicity in fish from regions with CTX-producing benthic dinoflagellates. Additionally, brevetoxins which are also Nav activators produced by the pelagic dinoflagellate *Karenia brevis,* have been detected in fish livers more than a year after algal blooms subside [32]. The CTX binding affinity for the primary Nav receptor was also shown to outcompete brevetoxin for the site [47]. Therefore, it is not unreasonable to consider CTXs have extremely long biological half-lives that are likely to vary based on the physicochemical properties of each toxin variant.

In this study of *L. rhomboides*, the calculated CTX burdens for each fish at the end of the depuration period were not measurably different than burdens measured (or estimated using the correlation between cumulative dose and total CTX burden) for day 20 (see Supplementary Figure S4). Although the level of uncertainty in the kinetic rate measurement limits the precision of our half-life estimate, the remaining high CTX burden after depuration suggests that the models presented here are valid and show that pseudoelimination via growth was a primary factor in lowering concentration in *L. rhomboides*. The growth correction Model 1, Model 2, and Model 3 (Figure 7B–D) showed that the use of mean and individual measured growth rate constants agreed, suggesting that this method is robust enough for use in food web modeling if accurate estimates of growth rates are available for wild-fish. Age and growth studies are available for numerous reef fish species globally and a thorough example of how these data could be useful was recently provided in a critical review on conceptual models for ciguatera risk assessment [23].

#### *3.6. Conclusions and Implications for Trophic Transfer*

In this study we report that C-CTX-1 was bioaccumulated in *L. rhomboides* from low environmentally relevant doses (0.02 ng CTX3C eq. g−<sup>1</sup> day−1), and present that total ciguatoxicity was retained for at least 3.3 months and depuration was mostly a function of fish growth, a pseudo-elimination factor. These data highlight that field determined concentrations of C-CTX may not reflect the mass balance of CTX that is bioavailable in food webs. Increased elimination may be expected in wild-caught fish that have varied diets and expend more energy, thus augmenting metabolic rates and lowering CTX levels. However, the retention of C-CTX-1 in *L. rhomboides* at a basal level is remarkable. This long-term CTX retention in fish helps to explain how regions with no recent sign of *Gambierdiscus* sp. blooms can produce toxic fish that cause CP outbreaks, whether by fish migration or after a period of low CTX production [12,48]. This study also suggests that field data may need to be staggered if *Gambierdiscus* abundance and toxicity are collected as a baseline in parallel with fish samples for food web toxin analysis, as is the traditional framework for field studies. Future studies could include repeated exposures with intermittent depuration periods to understand CTX kinetics once fish are re-exposed to the toxins while still depurating, which likely occurs in migratory fish and in areas with regular temporal swings in toxicity [6,9]. Quantifying the kinetics of elimination via respiration, feces, and metabolic transformations, as well as sequestration in reproductive tissues, may also be useful for future food web models of CTX exposure.

Improving our understanding of CTX kinetics has clear advantages to our interpretation of CTX dynamics in the field, and in developing improved risk models to prevent CP. Our prior efforts have shown clear pulses of C-CTX in the benthic algal community [9] in hyperendemic regions that may be dictated by lower temperature tolerance of toxigenic *Gambierdiscus* species [8,49]. If the timing of the CTX source pulses occur in higher frequency than the depuration of CTXs at higher trophic levels, we may expect increasing prevalence of toxic fishes once CTX-producing algae are established in a region. While more evidence on these dynamic cascades and factors that might influence them is needed, experimental studies such as these help to parameterize the capacity of toxin removal from the organism. From a resource management perspective, understanding periods of high and low risk, even if only possible for fish with high site fidelity, would be a great advantage to the fishery, especially when coupled to spatial resolution of CTX.

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

#### *4.1. Reagents and Chemicals*

All solvents were HPLC grade obtained from Fisher Scientific (Waltham, MA, USA) and were acetone, MeOH, *n*-hexanes, CHCl3, and dimethyl sulfoxide (DMSO). Bond Elut silica solid-phase extraction (SPE) cartridges (100 mg and 500 mg) were from Agilent Technologies (Santa Clara, CA, USA). Ouabain octahydrate (O) and veratrine hydrochloride (V) used for in vitro assays were from Sigma (St. Louis, MO, USA). The 3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyl- tetrazolium (MTT) was sourced from Alfa Aesar (Haverhill, MA, USA) and was prepared in sterile phosphate buffered saline (PBS) from Medicago (Quebec City, QC, Canada). Ciguatoxin 3C (CTX3C) was purchased from Wako Chemicals (Osaka, Japan) and a 50 ng mL−<sup>1</sup> stock was prepared in LCMS grade MeOH and aliquoted in sealed amber vials maintained at −20 ◦C until use.

Adherent murine neuroblastoma cells (Neuro-2a; ATCC CCL-131) were originally purchased from the American Tissue Culture Collection (Manassas, VA, USA) and modified (i.e., OV desensitized) lines generated and maintained to ensure maximal and stable cell response to CTX in MTT-based assays prior to use (available on request). Powdered Roswell Park Memorial Institute (RPMI) 1640 medium (Millipore Sigma, Burlington, MA, USA) was prepared in 10 L batches with sterile ultrapure water (18 mΩ) and 1L aliquots filtered (polyethersulfone 0.2 μM Supor membrane; Pall Corp; Port Washington, NY, USA) into sterile bottles. Supplements included sterile L-glutamine (200 mM stock), sodium pyruvate (100 mM stock), and fetal bovine serum (all from Gibco; Grand Island, NY, USA). Trypsin-EDTA (0.025% stock) used in cell detachment and harvest was from Corning (Corning, NY, USA). Cell culture consumables including serological pipettes, tubes, flasks, and micro-well plates were from CellTreat (Shirley, MA, USA). Trypan blue (Fisher Scientific, Waltham, MA, USA) was prepared to 0.2% in sterile PBS (pH 7.4) and used in cell enumeration and viability assessment.

#### *4.2. Controlled Exposure*

#### 4.2.1. Fish Collection and Acclimation

Wild pinfish, *Lagodon rhomboides* (Sparidae), were collected from Mississippi Sound (Dauphin Island, AL, USA) and Perdido Bay (Orange Beach, AL, USA) using hook-andline with sabiki rigs. To reduce stress from handling, fish were directly transferred to an aerated, seawater-filled cooler using a specialized dehooking tool which does not require fish handling. Fish were transported in aerated, temperature and salinity-controlled tanks to the experimental wet-lab facility at the Dauphin Island Sea Lab (DISL) within one hour. Prior to transfer to a primary acclimation tank, a five-minute freshwater dip was performed to remove marine ectoparasites. This step was important to reduce the chance of disease in the closed system. The recirculating aquarium system was composed of a 450 L sump, 375 L head tank, and 120 L enclosure tank and filled with artificial seawater (Crystal Sea Marine mix; Mount Dora, FL, USA). To limit stress, environmental enrichment provided to each tank habitat included a two-inch bed of pool filter grade silica sand, 12 cm diameter × 12 cm length PVC tubes for structure, and artificial submerged aquatic vegetation. Fish were collectively monitored in the acclimation tanks in groups of 25 for three weeks for signs of parasitic diseases and behavioral abnormalities prior to transfer into individual tanks for experimental treatment. Two fish collections were performed to supply fish for the exposure study, once in January 2019 and again in May 2020. The water conditions across all systems were maintained on a 12 h light: 12 h dark cycle with weekly water changes (artificial seawater: 15–17 psu; 24 ± 2 ◦C; ammonia ≤ 0.1 ppm, nitrite ≤ 0.1 ppm, nitrate ≤ 40 ppm, pH = 8.1 ± 0.1).

#### 4.2.2. Feed Formulation

Experimental (CTX feed) and control diets (no CTX) were created from a blend of fish meal, Mazuri Aquatic Gel Diet for Omnivorous Fish (sku: 1815252-409; Mazuri Exotic Animal Nutrition, St. Louis, MO, USA), and water. The fish meal consisted of white muscle tissue from either non-toxic great barracuda, *Sphyraena barracuda,* collected from the northern Gulf of Mexico, Alabama (control) or *S. barracuda* with naturally incurred C-CTX that were collected from the U.S. Virgin Islands during concurrent efforts. Fillets with skin, scales, and bones removed, were cut into small chunks and subsequently homogenized in an industrial food-grade stainless steel grinder (STX Turboforce 3000; Lincoln, NE, USA). From each fish, at least five replicate subsamples of minced *S. barracuda* tissue were extracted and analyzed for the presence of CTX-like activity using N2a-MTT, and C-CTX-1 confirmed by liquid chromatography-mass spectrometry as previously reported [50]. After verifying control and CTX fish, batches were pooled and mixed three times, then frozen in stainless steel trays and freeze dried. Dehydrated tissue (300 g) was powdered using a grain mill (HC-300; C-Goldenwall, Amazon, Seattle, WA, USA) at 28,000 rpm for 1 min and sieved to ≤500 μm to create a fine and uniform product. All fish powder was thoroughly mixed and kept in 500 g aliquots at −20 ◦C in airtight containers until use. Subsamples of the homogenized freeze-dried fish powder were taken for analysis by the N2a-MTT for final quantification of toxicity of the pelleted diet prior to further preparation. The control and CTX diets for *L. rhomboides* were created by combining equal parts Mazuri gel powder and fish powder, then homogenized and combined with 60% water (by weight) that was warmed to 65 ◦C. The homogenized feed was transferred to a pastry piping bag with a round tip (Wilton size 10), then dispensed in even rows onto pre-weighed parchment paper. After fully dispensing the gel, wet weight was recorded, and the gel was cut into 5 mm pellets. The tray was then placed in a drying oven at 65 ◦C for approximately 1 h to remove approximately 2/3 water so that pellets retained a uniform shape. To maintain a consistent product across batches, the water content (as a percent of the final product) was calculated by subtracting the weight of solid ingredients in the mixture from the weight of the formulated food after drying, then dividing by the whole-food dry weight. Subsamples of pellets were weighed on an analytical balance to collect an average pellet weight.

#### 4.2.3. Experimental Design

To accomplish optimal control and observation of fish, *L. rhomboides* were individually transferred to 12 L tanks attached to a closed recirculating aquarium system. Four recirculating aquarium systems with 10 to 15 tanks per system were used in this study. Fish were acclimated to the individual tanks (one fish per tank) for at least 2 weeks prior to the beginning of experimental treatment to ensure water quality parameters and fish health were stable. This acclimation period was in addition to the post-collection acclimation.

The experiments were designed to provide control over dietary intake of C-CTX-1 in individual fish so that the toxicokinetics of bioaccumulation and depuration could be analyzed across multiple tissues with sufficient replication for the subsequent statistical analyses. Tanks designated to receive control or CTX pellets were assigned a sampling day on the day prior to the start of experiments using a random number generator. Control fish received control pellets throughout trials while CTX fish received CTX pellets for up to 20 days (bioaccumulation phase) and then transitioned to the control feed for the depuration phase (see Table 1 for sampling scheme). Both control and CTX pellets contained powderized *S. barracuda* flesh (non-toxic or toxic, respectively) to match the nutritional intake across groups and maintain consistency (minus CTX) when fish were transitioned from the bioaccumulation to the depuration phase. Fish fed only the control pellets were designated for each sampling point to be used as matrix controls in assays, and to compare differences in behavior, and growth rate compared to the CTX treatment groups.

Data were collected from two feeding experiments (trial 1 and 2, hereafter) to allow adequate replication. The first experiment had maximum bioaccumulation and depuration phases of 20 days each with a total of 52 fish and was performed 5 March–13 April 2019. Fish were sampled as baseline controls prior to feeding on the first experimental day and on days 6, 10, and 20 of bioaccumulation and days 5, 10, and 20 of depuration. In the second experiment performed 30 July–25 November 2020, the bioaccumulation phase was replicated identical to trial 1, and depuration was extended out to a maximum of 99 days to increase the depuration course. Fish were collected on day 20 of the bioaccumulation phase to compare to trial 1 for reproducibility, and then on days 40, 70, and 99 of the depuration phase. Based on the lack of overt signs of intoxication during the first experiment and ample analytical controls, during the second experiment, remaining *L. rhomboides* were allocated towards maintaining experimental replication in the CTX treatment group (CTX fish, *n* = 4) and control replicates were reduced to *n* = 2 per time point (except day 0, *n* = 3) (Table 1).

Daily feed requirements were calculated based on initial weights (g) of fish which were recorded using a water displacement method one or two days prior to the beginning of exposure trials. All fish were fed either control or CTX pellet food at the same time daily, normalized to body mass (approximately 1.8% of initial body weight day−1). Fish were fed the same amount of food daily relative to the initial whole wet weight for the entire experiment. Individuals were closely observed to ensure all food was eaten during feeding times, and after acclimation *L. rhomboides* were consistently consuming the pellets within seconds of entering the tank. Fish were also re-weighed at multiple points (at minimum initial and at time of sampling) throughout the study to track growth rate and to determine if somatic growth influenced C-CTX concentration as previously expected by others in P-CTX fish studies [20,33,51].

#### 4.2.4. Fish Sampling, Dissection, and Extraction

On a designated day and time of sampling, fish were fed and left undisturbed for 7 h. Fish were euthanized according to approved IACUC protocols by iced seawater immersion and mortality was confirmed by cessation for at least 10 min. A secondary spinal transection was performed to ensure mortality, and fish were then dissected. Dissected tissues were weighed whole including muscle, brain, liver, gonads (when present), and additional visceral organs combined (heart, spleen, pancreas, gall bladder, intestine). Swim bladder was discarded and not analyzed in this study. Stomach was removed from the viscera samples to limit residual CTX signal from undigested food in downstream toxicity assays; however, given reported gut clearance rates <24 h, the amount of residual CTX in the intestinal tract was expected to be below the limit of quantification (LOQ; described in Section 4.3 Data Analysis) in the viscera on our assays.

Muscle subsamples (5–7 g) were extracted twice in acetone (2 mL g−<sup>1</sup> tissue weight) by bead disruption (2.6 mm diameter; ceramic) using a Bead Ruptor 24 (Omni International; Kennesaw, GA) at a speed of 5 m/s for two cycles of 30 s duration each. Resultant homogenates were centrifuged at 2465× *g* for 5 min at room temperature (approximately 21 ◦C) between extractions to obtain the supernatant. Combined supernatants were placed at −20 ◦C for 18 h, then centrifuged (4 ◦C, 2465× *g*, 10 min), and supernatants dried under a gentle nitrogen stream (45 ◦C). Dried residues were reconstituted in 90% aq. MeOH (1 mL g−<sup>1</sup> original weight) and partitioned twice using *n*-hexane (2 mL g−<sup>1</sup> original weight) to remove non-polar lipids. The aq. MeOH phase was dried under nitrogen, and the resultant residue was partitioned with CHCl3: H2O (50:50, v:v). The CHCl3 layer containing CTX was collected, then water phase partitioned again with the same volume of CHCl3. Pooled CHCl3 fractions were dried, reconstituted in 500 μL CHCl3, and further cleaned by silica SPE. Powderized fish tissue (CTX and control) was extracted in a similar manner using a 5:1 solvent (mL) to tissue (g) weight ratio. Non-muscular tissues were extracted whole due to their small size and solvent ratios adjusted accordingly. Final tissue extracts were dissolved in 1 mL 100% MeOH and stored at −20 ◦C until analysis.

#### *4.3. Toxin Analysis*

#### 4.3.1. Maintenance of Neuroblastoma Cells

Cells were maintained in vented 175 cm<sup>2</sup> sterile culture flasks with RPMI medium supplemented with 5% heat-inactivated FBS, 1 mM sodium pyruvate, and 2 mM L-glutamine (complete media) and maintained in a humidified water jacketed incubator at 37 ◦C with 5% CO2: 95% atmospheric air. Cells were passaged every 48 h and maintained in exponential growth. During passaging and in preparation for seeding 96-well plates, cells were harvested with 0.025% trypsin-EDTA for <2 min, trypsin deactivated with 10% FBS-RPMI complete medium, centrifuged, supernatant discarded and cells washed two times in PBS. Duplicate aliquots of cells resuspended in 5% FBS-RPMI medium were enumerated on a hemocytometer to calculate growth rates, and passage % viability via trypan blue staining.

#### 4.3.2. Neuroblastoma MTT Assay (N2a-MTT)

Sample extracts were tested for composite voltage-gated sodium channel (NaV) response using a standardized in vitro mouse neuroblastoma assay (N2a-MTT) as previously described [50] with toxin quantification of serially diluted samples evaluated and compared to a 9-point 2-fold serial dilution of CTX3C (initial dose equal to 20 pg well−<sup>1</sup> or 86.96 pg mL<sup>−</sup>1). The European Food Safety Authority (EFSA) has developed guidelines outlining toxic equivalency factors for the various CTX-group toxins [52]. CTX3C has been reported to be two-fold more toxic than C-CTX-1 based on intraperitoneal toxicity and was accepted here as the better certified reference standard compared to P-CTX1 which was reportedly ten-fold more toxic than C-CTX-1. The difference in CTX3C standard toxicity could cause lower estimates of C-CTX-1 content. For better understanding of toxin content, conversion from CTX3C eq. to C-CTX-1 can be done by multiplying CTX3C eq. concentrations by a factor of 2.

Due to the sample size of subsampled fish tissues, we were unable to perform additional LC-MS/MS analysis which requires much higher CTX concentrations compared to N2a-MTT. These analyses could have been accomplished by pooling tissues and/or extracts from replicate fish for each sampling point as described by others [20] but this would have lost the experimental replication that we deemed critical to the validity of this study, so was delayed to a future study with larger fish specimens.

N2a cells from established OV adapted lines, were seeded into 96-well plates at a density of 3 × 105 cells per well, in complete RPMI media (200 <sup>μ</sup>L). After 20 h, cells were dosed in triplicate with standards, controls, and sample extracts; all with and without OV. To prepare fish extracts for dosing, an extract aliquot dissolved in MeOH was transferred to a 1.5 mL microcentrifuge tube, dried under ultrapure N2(g), and redissolved in 5% FBS-RPMI complete media by vortex (30 s, room temperature). Cells in assay wells were carefully inspected by light microscopy prior to dosing and development. Positive controls (CTX-positive reference material), negative controls (containing PBS and medium), and assay controls with (sensitized) and without (non-sensitized) O/V (final well concentration: 0.22 mM ouabain/0.022 mM veratrine) were used to ensure quality assurance and control throughout the several hundred assays performed during this study. Sample wells were dosed with 10 μL of the fish extract solubilized in culture media or PBS (final well volume 230 μL). After a 20 h incubation, well contents were removed and MTT (1 mg/mL) diluted in 5% FBS-RPMI-1640 complete medium was added for 30 min. The resultant insoluble formazan product produced by mitochondrial activity of remaining live cells was solubilized in 100% DMSO (100 μL) with the colorimetric change measured within 5 min on a spectrophotometric microplate reader (μQuant; Biotek Instruments; Winooski, VT, USA) at 570 nm. Cells sensitized with O/V were used to assess CTX-like activity, while non-O/V-sensitized cells were used to monitor non-specific activity induced by sample extracts. When O/V-dependent toxicity was detected by at least a 20% difference between controls and wells dosed with fish extracts, a two-fold serial dilution of extract was prepared and assayed parallel to a CTX3C standard dilution series on the same day.

#### 4.3.3. N2a-MTT Data Analysis

Raw data were analyzed using Microsoft Excel version 2013 (Microsoft Corporation, Redmond, WA, USA). Normalized data were analyzed with GraphPad Prism version 9.0.0 (GraphPad Software, San Diego, CA, USA).

Raw absorbance values for wells dosed with standards and fish extracts were normalized to the OV control wells (20 μL O/V + 10 μL PBS, or 30 μL PBS), to account for minor OV N2a mortality established during cell line adaptation as described by others [53]. Triplicate absorbance responses within a single assay plate were deemed acceptable when the relative standard deviation was below 20%. To produce standard curves for quantification on the day of each assay, CTX3C standard doses (x-values) were logarithm transformed and fit by non-linear regression against the normalized response (y-values) in wells at each dose using a four-parameter logistic equation with variable slope (Y = Bottom + (Top/Bottom)/(1 + 10ˆ((LogIC50 − X) ∗ HillSlope). Toxicity of tissue extracts were estimated by interpolating the normalized responses (y-value) in wells onto the standard curve to estimate the dose (unknown x-value) in pg CTX3C equivalence. Values that fell between the effect concentration 20–80 (EC20 to EC80) which is the linear portion of the standard curve are acceptable, but we chose to use more strict parameters for quantification by accepting only values between EC30 and EC75 because full curves with a top and bottom plateau were not always possible based on the CTX concentrations in tissues. Interpolated results (pg CTX3C eq.) were divided by the tissue equivalence (mg TE) of the dose to calculate the concentration in ng CTX3C eq. g−<sup>1</sup> TE. The LOQ for CTX in each tissue type was determined by dividing the mean value of the EC75 from the CTX3C dose–response curve and the maximum TE dosed in wells without a matrix induced effect defined as either growth enhancement >20% of the control wells or cell death associated with non-Nav mechanisms (evaluated in sample wells without OV). All samples were analyzed in triplicate across 2–4 independent assays performed on separate days with a CTX3C standard curve prepared on the day of each assay.

#### *4.4. Ciguatoxin Kinetics*

#### 4.4.1. Muscle, Liver, and Viscera Ciguatoxin Kinetics

Kinetics of CTX uptake and depuration in separate tissues were investigated through non-linear regression analysis of the experimentally determined CTX concentration in sample extracts against time for both phases using GraphPad Prism version 9.0.0 (model comparison function). When CTX was below detection levels due to a non-specific matrix effect, those replicates were excluded from analyses (see Results Section 2.5). Normality of toxin data was evaluated using a Shapiro–Wilk test and models were fit using a least squares regression with no weighting and compared by an iterative process. Best fit models were compared based on the Akaike's Information Criterion (AICc) which balances the goodness of fit using sum-of-squares and the simplicity (number of degrees of freedom) of the two models.

#### 4.4.2. One-Compartment Model Kinetics

The combined CTX concentrations (*Cfish*) for the measured tissues were calculated for each fish by multiplying the quantified concentration for each tissue (muscle, liver, and viscera) by the whole tissue mass (conc. x mass = tissue burden), summing the tissue CTX burdens, and dividing by the total sum of the whole muscle, liver, and viscera mass. The *Cfish* was used to investigate a one-compartment model of CTX kinetics which assumes a homogenous concentration in the fish. While our analyses do not account for CTX in some compartments (kidney, stomach, and carcass), the included tissues (muscle, liver, viscera) contained the majority portion of CTX in the major tissue compartments in the fish. No CTX activity was detectable in brain or gonads, so these were excluded.

The uptake rate was calculated by linear regression of the measured *Cfish* (ng g−1) during the bioaccumulation phase against time (d = 0 to 20)

$$C\_{fish} = k\_{uptake} \times t + a \tag{1}$$

where the slope of regression is equal to the rate of increasing CTX concentration (*kuptake*; ng g−<sup>1</sup> day<sup>−</sup>1), *t* is time (d), and *a* is a constant which in this case is the y-intercept (ng g<sup>−</sup>1).

The first-order depuration rate constant (*k*2) was calculated using the methods outlined by Brooke and Crookes [45]. A one-compartment model of exponential decay was fitted to the measured *Cfish* during the depuration phase

$$\mathcal{C}\_{fish(t)} = \mathcal{C}\_{fish(i)} \ e^{-k\_2 t} \tag{2}$$

where *Cfish(t)* is the concentration (ng g−1) measured at the time of sampling, *Cfish(i)* is the initial concentration (ng g−1) at the start of the depuration phase (day 20), *t* is time (d), and *k*<sup>2</sup> is the overall depuration rate constant (day−1). For curve fitting, the measured *Cfish* in the depuration phase were natural Log transformed, i.e., Ln [*Cfish*], to allow linear regression of Log-concentrations versus time in which the slope of regression is the *k*2, and the coefficient of determination is used to confirm first-order kinetics. The *k*<sup>2</sup> calculated by these methods is the overall elimination rate constant including the sum of four first-order kinetic processes

$$k\_2 = k\_r + k\_m + k\_{\mathfrak{e}} + k\_{\mathfrak{X}} \tag{3}$$

where *kr*, *km*, and *ke* (all in units day−1) are the rate constant for elimination via respiration, metabolic transformation, and feces, respectively, and *kg* is the rate constant for the change in concentration due to fish growth, a pseudo-elimination process. For an indepth summary on each of the rate constants and application to overall *k*2, see Gobas [54] (pp. 4–9).

To correct the overall *k*<sup>2</sup> for growth of the fish, *kg* was calculated using the growth rate data collected for each fish. Fish masses were applied to an exponential growth model

$$\mathcal{W}\_{fish(t)} = \mathcal{W}\_{fish(i)} \; e^{k\_3 t} \tag{4}$$

where *Wfish(t)* is the wet weight (g) of the fish at any point, *Wfish(i)* is the initial wet weight at the start of the experiment, and *kg* is the growth rate constant (day<sup>−</sup>1). To allow a linear fit, the inverse of measured weights was natural Log transformed, i.e., Ln (1/*Wfish*), and plotted against time where the slope is equal to *kg* (day−1). The effect of growth, which results in diluted concentrations over time, was factored out of the depuration rate constant by subtracting the *kg* from the overall *k*<sup>2</sup> to give the growth-corrected depuration rate constant (*k*<sup>2</sup> *growth-corrected*; day<sup>−</sup>1)

$$k\_{2\\_growth-corrected} = k\_2 - k\_{\mathcal{S}} \tag{5}$$

and describes the rate constant for elimination processes that result in removal of CTX from the fish (*kr* + *km* + *ke*) [45].

#### 4.4.3. Kinetic Modeling and Correction of Growth Dilution

The exponential growth Equation (4) was used to estimate the mass of each fish on day 20 (end of bioaccumulation phase) that entered the depuration phase using measured growth rate constants (*kg*) and solving for *Wf(t)*. For trials 1 and 2, respectively, values of *t* were set to 20 and 22 days, because initial weights were collected on day 0 (trial 1) and two days prior to start (trial 2), while depuration was initiated on day 20 for both trials. The sum of muscle, liver, and viscera mass was estimated at day 20 using a linear correlation between whole fish mass (g) and the sum of tissue masses, i.e., Σ (whole muscle + liver + viscera; g), that were dissected from sampled fish for CTX analysis. The combined CTX burden in

the whole muscle, liver, and viscera at day 20 was estimated from the linear relationship between the cumulative dose administered (total ng CTX3C eq. consumed = number of pellets fed daily × number of days in bioaccumulation phase × pellet CTX3C eq. conc.) and the combined CTX burden measured in fish that were sampled between 0 and 20 d. Estimates of total CTX burden and combined tissue mass at the end of the bioaccumulation phase (day 20) were divided for an estimate of an overall CTX concentration in the fish at the start of depuration (*Cfish(i)*).

A simulation model of our data was created using the estimated *Cfish(i)* to study the effect of fish growth on elimination of CTX. Final concentrations were simulated for each fish in the depuration phase at all time points using the one-phase exponential decay model (Equation (2)) where *Cfish(i)* is an estimate on day 20, time is delta time in depuration depending on when the fish was sampled, and overall *k*<sup>2</sup> was measured from the linear plot depuration data. The resulting simulated concentrations were natural Log transformed (as previously described for the measured *Cfish*) and plotted against time (d) to compare the regression slope (*k*2) with the slope from the measured data. A correlation was then performed to check how well the simulated data fit the measured concentrations.

Simulated concentration data served as the baseline for growth-corrected models to compare the change in half-life (Ln [2]/ *k*) under different conditions of *k2 growth-corrected*. The growth-corrected models were produced by adjusting the depuration rate constant (*k*2) in the simulation exponential decay model. Three growth-corrected models were produced using different values of *k*<sup>2</sup> *growth-corrected* that were calculated based on fish growth rates in the depuration phase, i.e., average growth rates at each sampled time point (Model 1), average growth rate based on ANOVA grouping (Model 2), and the individually measured growth rate (Model 3). Results of each model compared to the simulation are presented using the linear version of the exponential decay model where concentrations are natural Log transformed and plotted against time. The growth-corrected half-lives of the models were finally compared to the half-life calculated from growth correction by a simpler approach reported by others [20,55] that uses a growth correction factor, where measured concentrations in fish during the depuration phase are multiplied by (1 + *kg* × time).

#### *4.5. Statistical Analyses*

Fish size (mass in g) at the initial time point across studies was analyzed using a non-parametric two-tailed Mann–Whitney U test since trial 1 weights were not normally distributed. Growth rate constants (*kg*) of control and treatment groups collected at the same time point were tested for normality using a Shapiro–Wilk test and compared for differences using a two-tailed *t*-test, respectively. Growth rate and toxicity data between control and exposure treatments were tested for normality and homoscedasticity using a Shapiro–Wilk test and Brown-Forsythe test, respectively, and residual plots were visually inspected as a parallel method prior to one-way ANOVA with Tukey's multiple comparisons test. Growth trends were investigated by a linear regression plot of *kg* against time. A simple correlation was used to analyze the fit of the measured and model simulated *Cfish(t)* in the depuration phase prior to investigating growth correction of the model.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/10 .3390/toxins13110774/s1; Supplementary Figure S1: Representative in vitro dose–response curves from the neuroblastoma-MTT assay in this study. Supplementary Table S1: Results of a *t*-test on the first order growth rate constants (*kg*) between Control and CTX-treated *L. rhomboides* at each sampled time point. Supplementary Figure S2: Results of ANOVA with Tukey's test on *kg* of Control and CTX feeding treatments across time. Supplementary Figure S3: Results of ANOVA with Tukey's test for C-CTX-1 measured in CTX3C eq. concentrations (mean ± s.d.) in muscle, liver, and pooled viscera of *L. rhomboides* in this study. Supplementary Figure S4: Sum of total C-CTX burdens (measured as CTX3C eq.) in the tissues of *L. rhomboides* sampled in this study.

**Author Contributions:** Conceptualization, C.T.B. and A.R.; Investigation, C.T.B.; Methodology, C.T.B. and A.R.; Analysis and Visualization, C.T.B. and A.R.; Writing—original draft, C.T.B.; Writingreview and editing, A.R.; Funding acquisition, A.R.; project administration, A.R.; supervision, A.R.; and resources, A.R. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was funded by the National Oceanic and Atmospheric Administration (NOAA NOS NCCOS) Ecology and Oceanography of Harmful Algal Blooms (ECOHAB) program (CiguaTOX: NA11NOS4780028) and is publication 985. Research and graduate student funding were also supported through the National Science Foundation (NSF) Partnerships in International Research and Education Program (CiguaPIRE; 1743802) and contributes to the NSF and NIEHS Center for Oceans and Human Health: Greater Caribbean Center for Ciguatera Research (NSF: 1841811; and NIH: 1P01ES028949-01). C.T.B. was supported by a graduate research assistantship funded by NOAA and NSF and received additional internal research support from the Bullard Fund (Marine Sciences, University of South Alabama).

**Institutional Review Board Statement:** The study was approved by the Institutional Animal Care and Use Committee (IACUC) Review Board of the University of South Alabama (protocol number 1105747-2 approved 24 August 2017, and 1640293-2 approved 27 August 2020).

**Informed Consent Statement:** Not applicable.

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

**Acknowledgments:** We are thankful to Matthew Boehm, Diana Marchant, and Jonathan Wittmann from the Dauphin Island Sea Lab (DISL) for additional support of aquarium and fish maintenance during this extended study. We also appreciate the aquarist expertise of Brian Jones (DISL Estuarium). Special thanks to Tony Dovi (Dovi Design & Illustration) for assistance with original preparation of graphical art used in the graphical abstract.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**


## *Review* **Advances in Detecting Ciguatoxins in Fish**

### **Tibor Pasinszki 1,\*, Jimaima Lako <sup>2</sup> and Todd E. Dennis <sup>3</sup>**


Received: 28 June 2020; Accepted: 26 July 2020; Published: 31 July 2020

**Abstract:** Ciguatera fish poisoning (CFP) is currently the most common marine biotoxin food poisoning worldwide, associated with human consumption of circumtropical fish and marine invertebrates that are contaminated with ciguatoxins. Ciguatoxins are very potent sodium-channel activator neurotoxins, that pose risks to human health at very low concentrations (>0.01 ng per g of fish flesh in the case of the most potent Pacific ciguatoxin). Symptoms of CFP are nonspecific and intoxication in humans is often misdiagnosed. Presently, there is no medically approved treatment of ciguatera. Therefore, to mitigate the risks of CFP, reliable detection of ciguatoxins prior to consumption of fish tissue is acutely needed, which requires application of highly sensitive and quantitative analytical tests. During the last century a number of methods have been developed to identify and quantify the concentration of ciguatoxins, including in vivo animal assays, cell-based assays, receptor binding assays, antibody-based immunoassays, electrochemical methods, and analytical techniques based on coupling of liquid chromatography with mass spectrometry. Development of these methods, their various advantages and limitations, as well as future challenges are discussed in this review.

**Keywords:** ciguatera; ciguatoxin; cytotoxicity assay; ELISA; HPLC; immunoassay; LC-MS/MS; mouse bioassay; receptor-binding assay

**Key Contribution:** Methods to detect and quantify ciguatoxins in fish tissue are critically reviewed.

### **1. Introduction**

Ciguatera fish poisoning (CFP), currently the most common marine biotoxin food poisoning worldwide, is a non-bacterial foodborne disease associated with consumption of circumtropical fish and marine invertebrates that are contaminated with polyether sodium channel activator neurotoxins (ciguatoxins, CTXs) [1–12]. CTXs are a family of heat-stable and lipid-soluble compounds that cannot be degraded by normal cooking. CTXs are colorless and odorless, therefore cannot be detected by smell or visual inspection of fish flesh. Ciguatoxins are produced by certain benthic dinoflagellate species from the *Gambierdiscus* and *Fukuyoa* genera and enter the marine food chain via herbivorous fish and invertebrates [2,5,9,13]. These toxins are subsequently biotransformed in herbivorous, omnivorous, and carnivorous fishes to more oxidized and more potent forms of CTXs and accumulate to toxic levels in edible fish. During the biotransformation of P-CTX-4B to P-CTX-1 (see below) there is a ten-fold increase in potency [8]. The structure of CTXs varies according to geographic distribution; therefore, they are classified as Pacific Ocean (P-CTX), Caribbean Sea (C-CTX) and Indian Ocean (I-CTX) ciguatoxins. P-CTX-1 is regarded as the most potent toxin, and the recommended safety limit for CTXs in fish for human consumption has been set at 0.01 ng P-CTX-1 toxin equivalent/g fish tissue (0.01 ppb

P-CTX-1 equivalent) by both the European Food Safety Authority (EFSA) and United States Food and Drug Administration (US FDA) [1,14]. The recommended safety level for C-CTX-1 equivalent toxicity is 0.10 ppb [1,9,14]. Although the safety limit for I-CTXs has not been published yet, based on experiments indicating that the toxicity of I-CTX-1 is 60% of that of P-CTX-1 potency [15], a safety level of 0.017 ppb for I-CTX-1 equivalent toxicity may be considered. CFP is known in tropical regions for centuries, and it is an increasing risk of food poisoning worldwide; it occurs now in non-endemic areas due to international trade of fish and fish products and the expansion of the geographic ranges of dinoflagellates as a likely result of global warming [12,16,17]. Intoxication by CTXs may cause neurological, gastrointestinal, and cardiovascular symptoms depending on the amount and type of the toxin ingested [1–7,18], and occasionally in severe cases, CFP can be fatal [19,20]. P-CTX-1 possesses risk to human health at concentrations higher than 0.022–0.1 ng g−<sup>1</sup> in fish flesh [8,21]. About 10,000 to 50,000 people suffer from the illness annually [1]; however, this is likely a substantial underestimate considering the incidence of non-reported cases from remote areas and non-diagnosis. Only 2–10% of CFP cases are estimated to be reported to health authorities [4]. Currently, there is no routine, rapid, reliable, and cost-effective point-of-care (POC) test that can detect ciguatoxins on-site or prior to consumption. Identification and quantification of CTXs is challenging even for laboratories due to the low CTX concentrations in fish flesh, the low recommended limit of 0.01 ng g<sup>−</sup>1, and the lack of reference materials and standards for all CTXs. The concentration of toxins in fish liver is about 10–50 times higher than in muscle tissue [22,23], thereby CFP becomes more problematic in communities consuming fish viscera. In extreme cases, such as for the liver of a large moray eel caught in Kiribati, toxicity can be as high as 539 ng g−1, 50,000 times higher than the accepted safety level of 0.01 ng g−<sup>1</sup> [23]. The symptoms of CFP were first described by Captain James Cook and Don Antonio Parra in the 17th century during their exploration of the Pacific Ocean and Caribbean Sea, respectively [3,24]. CFP was finally linked to dinoflagellates in 1977 [25]. Several methods have been developed to test for CTX presence in fish, ranging from indigenous observations and animal mortality tests to modern analytical techniques. The present review aims to summarize such methods and identify future challenges in CFP testing. *Gambierdiscus* strains, from which *G. polynesiensis* in the Pacific Ocean and *G. excentricus* in the Atlantic Ocean represent the major threat to human health [26,27], are known to produce not only CTXs but other toxins, such as the water-soluble and structurally related maitotoxins [28,29] and gambierones [30,31]. However, the contributions of these latter toxins to CFP is insignificant compared to that of CTXs, due to their high water solubility and low oral potency [8,9]; therefore, these toxins are beyond the scope of this review.

#### **2. Ciguatoxins**

The metabolic modification of dinoflagellate toxins in fish produces a large number of structurally related CTX congeners. Multiple CTX congeners exist in fishes, and each may contribute to CFP. To date, 47 CTXs have been identified but less than half are structurally characterized due to the insufficient amounts of pure toxin available for analysis. Legrand et al. isolated 0.35 mg of pure P-CTX-1 from 125 kg of fish viscera, including 43 kg of liver, from 4150 kg of moray eels, *Gymnothorax javanicus* [32]. The specific chemical structures of major CTXs, however, were elucidated using NMR and mass spectrometry [33–42]. CTXs are composed of contiguous cyclic ether rings aligned in a ladder-like fashion, and the two termini of the rigid ladder are varied in congeners. Most of the CTX congeners possess a primary hydroxyl group that may allow selective derivatization. The toxicity of various CTX congeners are different. On the basis on their acute intraperitonial median lethal dose (LD50) in mice, EFSA has adopted the following toxicity equivalency factors (TEFs) for CTXs: P-CTX-1 = 1, P-CTX-2 = 0.3, P-CTX-3 = 0.3, P-CTX-3C = 0.2, 2,3-dihydroxy-P-CTX-3C = 0.1, 51-hydroxy-P-CTX-3C = 1, P-CTX-4A = 0.1, P-CTX-4B = 0.05, C-CTX-1 = 0.1 and C-CTX-2 = 0.3 [1].

To date twenty-two CTXs have been identified from Pacific fish samples (Table 1). The skeletal structures of the 22 structurally characterized toxins can be separated into two groups, the P-CTX-1 (or CTX-1B) type and the P-CTX-3C type. P-CTX-1 (mass 1110.6 Da, C60H86O19) exhibits the highest

toxicity against mice [33]. Molecular masses of P-CTXs are summarized in Table 1, and structures of CTXs are shown in Figure 1.


**Table 1.** Molecular formula and mass (in Da) of identified ciguatoxins 1.

<sup>1</sup> OH = hydroxy, H = hydro, n.a. = not available; <sup>2</sup> Epimers; <sup>3</sup> Epimers; <sup>4</sup> Epimers; <sup>5</sup> Epimers. Alternative or old names: P-CTX-1 = CTX-1B and CTX; P-CTX-2 = 52-*epi*-54-deoxy-CTX-1B; P-CTX-3 = 54-deoxy-CTX-1B; P-CTX-3B = 49-*epi*-P-CTX-3C; P-CTX-4B = 52-*epi*-P-CTX-4A, GTX-4B, GT-4B or gambiertoxin-4B; 49-*epi*-P-CTX-3C = P-CTX-3B; 56-*epi*-C-CTX-1 = C-CTX-2; 2,3-dihydro-2,3-dihydroxy-P-CTX-3C = 2,3-dihydroxy-P-CTX-3C = CTX-2A1.

Twelve Caribbean CTXs have been identified thus far [34], and the structures of the two major toxins (epimers C-CTX-1 and C-CTX-2, mass 1040.6 Da, C62H92O19, see Figure 2) have been determined [35]. The molecular structure of C-CTX-1 has been recently revised [36] (Figure 2). Based on molecular fragmentation in a mass spectrometer at high collision energies, the N-ring of C-CTX-1 is more likely to be a seven-membered ring [36] than a six-membered [35]. Structure and toxicity of the other 10 CTXs have not been established yet. C-CTX-1 is considered to be 10-times less toxic than P-CTX-1 [8].

Six CTXs have been identified to date from fishes and sharks of the Indian Ocean [15,19,43], two isomer pairs with masses of 1140.6 Da (I-CTX-1 and -2, C62H92O19) and 1156.6 Da (I-CTX-3 and -4, C62H92O20) [15], as well as two congeners with only 2H less, 1138.6 Da (I-CTX-5, C62H90O19) and 1154.6 Da (I-CTX-6, C62H90O20), which corresponds to the formation of a double bond [19]. The exact structure of these toxins has not been determined as yet. Various experiments have indicated that the toxicity of both I-CTX-1 and -2 is 60% and both I-CTX-3 and -4 is 20% of the P-CTX-1 potency [15].

**Figure 1.** Structure of Pacific CTXs (alternative or old names are given in Table 1).

**Figure 2.** Structure of C-CTX-1 and C-CTX-2, according to Lewis et al. [35], and the suggested new structure, C-CTX-1(new) [36].

#### **3. Extraction of Ciguatoxins from Fish Tissue**

Extracting CTXs from fish tissue is a critical step in CTX quantification and strongly influences analyte recovery and thus analytical reliability. This purification step is also important to efficiently remove matrix-derived interfering compounds that negatively affect sample analysis, such as lipids. Extraction methods involve multiple steps, and are time consuming; they include, in general, the following major steps: (1) extraction of raw, freeze-dried or cooked muscle tissue with a polar organic solvent (typically acetone or methanol); (2) purification of the extract by liquid-liquid partitioning

(using diethyl ether, chloroform, or dichloromethane); (3) defeating the extract by liquid-liquid partitioning using hexane or cyclohexane; and (4) purification of the crude extract by solid-phase extraction (SPE), in one step using normal-phase or reverse-phase SPE, or more typically in two steps using consecutive orthogonal SPE phases. Several modified versions of the original extraction method of Lewis et al. [44] have been published during the last three decades, e.g., [12,43,45–49]; two currently used methods are summarized in Figure 3 as examples [41,45,50,51]. Varying the solvent used for extraction, the sample-to-solvent ratio, the number of extraction cycles used to extract fish tissue, and the number of SPE steps can influence the extraction efficiency of CTXs and extract purity, and determine time for extract preparation. A further complicating factor is that extraction efficiency depends also on the CTX analogue; for example, methanol is a good extraction solvent for various CTXs, but produces high levels of co-extractives. Using more polar solvent such as aqueous methanol limits the amount of co-extractives in the extract, is effective for the more polar CTX analogues, but much less effective for less polar CTX analogues [49]. Application of consecutive purification steps varies considerably among sample preparation protocols that exist in the literature. Most of the existing protocols are reviewed by Harwood et al. in 2017 [49]. Selection of the extraction protocol is not unambiguous, and unsurprisingly no validated extraction method exists to date. Current extraction protocols are complicated and not efficient enough.

**Figure 3.** Examples of extraction of CTXs from fish tissue for various CTX tests [41,45,50,51].

*Toxins* **2020**, *12*, 267

The fish tissue extraction process due to several purification steps is slow compared to the time frame of modern analytical techniques (e.g., liquid chromatography-tandem mass spectrometry, LC-MS/MS, see below), therefore it is the rate-determining step for testing a fish sample. To decrease extraction time, a ciguatoxin rapid-extraction method (CREM), which uses only 2 g of fish tissue and combines the first three extraction steps mentioned above by applying a methanol-hexane mixture, has also been developed by Lewis et al. [52]. The method was updated and modified later by Stewart et al. [53] and Meyer at al. [54]. The rapid extraction method was estimated to be two-to-three times faster than the standard method for extraction and clean-up of CTXs [53]. It is, however, less effective than 'normal' protocols in terms of efficiency and toxin yield [54], and therefore not widely used by all specialized laboratories. Although, in principle, higher extract purity is better for all CTX detection techniques, it is worth considering minimum extract purity requirements for various analytical techniques, regarding time and cost. The crude extract is sufficient for mouse bioassay (MBA), the SPE provides sufficiently pure extract for Cell-based Assay (CBA), Receptor-binding Assays (RBA), Enzyme-Linked ImmunoSorbent Assay (ELISA) and LC-MS/MS. Further high-performance liquid chromatography (HPLC) separation produces purer fractions which can be useful for specific investigations, such as for example the determination of toxicity profile using CBA-N2a (see Section 4.4 below).

#### **4. Detection and Quantification of Ciguatoxins**

Due to the serious threat to human health caused by CFP, to date a wide variety of methods have been developed to detect CTXs in fish, including native tests, animal mortality tests, biological methods (cytotoxicity assays, receptor-binding assays and immunoassays), and chemical methods (HPLC with fluorescence detection, LC-MS/MS) (Figure 4). Many of these methods are not specific to CTXs, inadequate for quantification, or allow quantification of CTXs only with results expressed in "equivalent of a CTX standard". Currently, the most advanced methods for monitoring CTXs are based on combination of biological and chemical methods into two steps by screening fish extract toxicity with sensitive functional assays first, followed by confirmation of the presence of CTXs via LC-MS/MS.

**Figure 4.** Timelines of CTX detection methods.

#### *4.1. Indigenous Tests*

Island communities that are strongly dependent on fish for food resources have developed various means over centuries to decrease the risk of CFP [10,11,55,56]; these include rubbing a small piece of liver on the mouth or skin and then testing for itchiness, cooking fish with a silver coin or copper wire and assessing discoloration, observing the color of fish gallbladder, examining food avoidance by ants and flies, feeding dogs, cats or pigs with suspected fish and observing sickness or fatality of animals, and bleeding and *rigor mortis* tests [10,11,55,56]. A fish is considered to be toxic in the bleeding test if haemorrhagic signs are visible at an incision on the tail of the dead fish. In the *rigor mortis* test, a fish is considered to be toxic if its flesh is flaccid an hour after death. All of these native and traditional test methods are now discredited due to their lack of specificity and the regular occurrence of both false negative and positive results. Darius et al. investigated the accuracy of the bleeding and *rigor mortis* tests by comparing test results with laboratory toxicity data obtained via the RBA and CBA on neuroblastoma cells [56], and concluded that intoxication in communities where CFP is highly prevalent may be reduced on the basis of traditional knowledge and a good understanding of high-risk versus relatively safe fishing areas.

#### *4.2. Animal-Feeding Bioassay Tests*

CTXs are toxic to a wide range of animal species [57]. Animal bioassay tests were developed during the 20th century. These tests were based on feeding cats [58], mongooses [59], chickens [60], or Dipteran larvae [61] with the flesh or viscera of suspected fish, or treating mosquitoes [62] or brine shrimp larvae [63] with fish extracts, and observing signs of intoxication and death of animals over time. Symptoms of cats and mongooses after being fed with toxic fish have been found to be similar in some respects to those of humans [59]. These reactions after a single test feeding, within 48 h, have been classified in five stages based on the maximal response of the test animal and numbered as: 0 = no symptoms; 1 = slight weakness and flexion of the forelimbs; 2 = slight motor ataxia, more pronounced flexion of the forelimbs, and weakness of the hind limbs; 3 = moderate motor ataxia with weakness and partial paralysis of limbs and body musculature; 4 = acute motor ataxia and extreme weakness or coma; and 5 = death. Stages 3–5 are indicative of high toxicity, 1–2 moderate toxicity, and 0 non-toxicity in fish. However, neither cats nor mongooses are satisfactory test animals to establish an LD50 because cats often regurgitate the test meal and mongooses commonly consume too much fish to permit the necessary replicate testing [59]. The chicken-feeding test is based on force feeding 8–10 days old chicks with cooked fish tissue (10% of body weight) and assessing the change in body weight of the study animals over a 48-h period [60]. The response of chickens to being fed contaminated fish liver has been found to be roughly quantitative. The mosquito bioassay test, where mosquitoes (*Aedes aegypti*) are intrathoracically injected, requires much smaller amounts of fish samples (8 g) than the tests discussed above, is much cheaper due to the low cost of mosquitoes, and is able to provide a LD50 value in2h[62]. This test requires, however, a fish extract, therefore use of laboratory, as it is not practical to conduct the extraction procedure under field conditions. Based on the mosquito bioassay and human symptomatology, the minimum lethal dose of P-CTX-1 in humans has been estimated to be 0.02 ng g−<sup>1</sup> [64]. Although brine shrimp are seemingly unaffected by consuming finely ground ciguateric fish, their larvae have been found to be sensitive to fish extract. The brine shrimp larvae bioassay was based on treating approximately 100 freshly hatched larvae in artificial sea water with fish extract and observing the proportion of larvae that died over a 24-h period [63]. The Dipteran larvae-feeding test is extremely simple, does not require cooking or any pretreatment of fish samples, and can be evaluated visually; therefore, this test has been suggested to be appropriate for use in communities inhabiting remote islands that have no laboratory facilities [61]. The larvae, however, are sensitive to other toxic substances. In the Dipteran larvae test, ten larvae are placed on ca. 5 g of fish sample and the inhibition of larval growth is followed for 3−24 h. Fish samples containing more than 1 ng CTXs in 1 g of flesh kill the larvae in about 3 h. By weighing the larvae and comparing them to healthy reference samples, a limit of detection (LOD) of 0.15 ng g−<sup>1</sup> can be achieved [61]. Although animal feeding tests, in general, are simple, easy to implement and do not require complex analytical equipment, such tests are insufficiently sensitive, incapable of providing specific information on individual toxins, time consuming, cannot be automated, and are expensive due to the required animal facilities and expertise. Further, there are serious ethical concerns about the application of these tests. Unsurprisingly, none of these tests currently are used in modern laboratories.

#### *4.3. Mouse Bioassay (MBA)*

The mouse bioassay [65,66] is the only animal test today that remains in use, e.g., [67], despite its disadvantages and ethical concerns. The MBA is simple and does not require complex analytical equipment, but it is expensive due to the need for animal facilities, is time consuming, and cannot be automated. The MBA provides information only about the total toxicity of a sample, therefore lacks specificity, and CFP caused by CTXs may be overestimated. The limit of quantitation (LOQ) of MBA is approximately 0.56 ng g−<sup>1</sup> for P-CTX-1 [1], therefore the bioassay is insufficiently sensitive to cover the suggested tolerance limit of 0.01 ng g−<sup>1</sup> in fish. CTXs are highly potent toxins in mice by either the intraperitoneal (i.p.) or oral route [33]. Injecting fish extract intraperitoneally into mice is the generally applied method in MBA [1,3]. Raw extracts are usually suspended in 1% Tween 60, 0.9% saline solution prior to injection. The test is administered either by establishing dose/survival-time relationships or by observing mice for motor ataxia or other bodily dysfunction for 24 h after injecting serial dilutions of CTX extracts. The end point in the assay is the death of the test animal. The dose-vs-time-to-death relationships for CTXs (Figure 5) are found to be as follows: log(dose) = c log(1 + *t* <sup>−</sup>1), where dose is in mouse units (MU), time (*t*) to death is in hours, and constants c is 3.3, 2.4, 3.9, and 2.3 for pure P-CTX-1, pure P-CTX-2, pure P-CTX-3, and partially purified P-CTX, respectively. One MU is defined as the i.p. LD50 dose for a 20 g mouse, equal to 5.0, 46 and 18 ng P-CTX-1, PCTX-2, and P-CTX-3, respectively [33].

**Figure 5.** Relationship between intraperitoneal dose in mice and time to death for Pacific CTXs, reproduced with permission from [33]. Copyright Elsevier Ltd., 1991. The equation of the fitted curves is shown in the main text.

#### *4.4. Cell-Based Assay (Cytotoxicity Assay)*

Cell-based assays are dependent on the toxic activity of fish extracts on cultured cells and reflect the combined potency of related toxins in the mixture. The overall toxicological effects of CTXs are caused by the action of CTXs on neuronal potassium and voltage-gated sodium channels [2]. CTXs bind quasi-irreversibly to voltage-sensitive sodium channels, enhancing sodium influx into cells, causing them to open at the normal cell-resting membrane potential, thereby impeding normal function. In addition, CTXs also inhibit neuronal potassium channels, which is likely to act in concert with effects on voltage-gated sodium channels to increase neuronal excitability [2,3]. Evaluation of cell viability forms the basis of CTX determination in CBAs [1].

Several cell and tissue-based assays were developed previously, namely the guinea pig ileum [68], guinea pig atrium [69,70], isolated frog nerve fiber [71], crayfish nerve cord [72], and blood cell hemolytic [73] tests. These tests, however, are outperformed by the mouse neuroblastoma cell assay (CBA-N2a) and thus are no longer employed. The CBA-N2a is widely used today, and its development by Manger et al. [74] is one of the most important milestones of replacing MBA in modern laboratories. The assay is based on the colorimetric detection of metabolically active N2a cells treated with CTX extract in the presence of ouabain/veratridine [74–76]; it expresses negatively the concentration of various voltage-gated sodium channel toxins and assesses cell death [3]. CTXs have no cytotoxic effect on N2a cells, therefore their detection requires addition of veratridine (a sodium-channel-activator that have a different binding site than CTXs) and ouabain (a sodium/potassium ion ATPase inhibitor). The combined effect of CTXs together with ouabain and veratridine causes an elevation of intracellular sodium ions to toxic levels in cells and a resultant decrease in cell viability that can be measured as a function of CTX concentration. Toxins are detected as a dose-dependent loss of cell viability, based on an end-point determination of mitochondrial dehydrogenase activity, due to the synergistic effect of ouabain/veratridine-induced cytotoxicity by CTXs (Figure 6a). Color development is based on the ability of active cells to reduce 3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyltetrazolium (MTT) to a blue-colored formazan product. The advantage of this MTT-based bioassay is that it is more sensitive for CTXs (at ng/g fish level) than the mouse bioassay, and suits to automation due to color reading [1]. Results obtained from CBA-N2a bioassays of fish extracts have correlated well with those obtained from MBA [1,74]. Although CBA-N2a have been widely used in the last three decades and the protocol has undergone numerous changes [77], a consensus assay protocol is still lacking. To this end, research to standardize this CBA [77,78] and to avoid matrix effects [79] also have been conducted. Viallon et al. revisited recently the CBA-N2a assay by investigating six key parameters, namely cell seeding densities, cell layer viability after 26 h growth, MTT incubation time, veratridine and ouabain treatment, and solvent and matrix effects [77]. A step-by-step protocol was defined by identifying five viability controls for the validation of CBA-N2a results, therefore, the improved method is an important step towards implementation of a reference detection test.

An advantage of the CBA-N2a assay is that necessary materials and reagents, as well as basic laboratory equipment are commercially available. A major disadvantage of CBA-N2a, however, is that it is time consuming, including, in general, 24 h incubation of neuro-2a cells, 24 h exposure of the neuro-2a cells to fish extracts, and a 4–6 h cell viability assessment. Achieved LOD and LOQ using this assay depends also on assay protocol. Nonetheless, CBA-N2a is able to provide LOD below the clinically relevant toxin levels in fish tissue. Selected examples for CBA application are summarized in Table 2 (more data are provided in the cited articles).

Fairey et al. modified the cell-based directed cytotoxicity assay and developed a reporter-gene modification by using CBA-N2a clones expressing c-*fos*–Luciferase; the assay was thereby utilizing a luciferase-catalyzed light generation as an endpoint and a microplate luminometer for quantification [81,82]. c-*fos* is an immediate response gene and a sensitive biomarker to localize the effects of toxins. This assay, however, is not commonly used possibly due to the problematic interpretation of the bell-shaped dose–response curve and the cost of fluorescent dye.

N2a cell lines are widely and routinely used in cell-based assays. However, other cell lines have also been tested for potential application in CBA. Zimmermann et al. [83] and Lewis et al. [84] developed fluorescent CTX assays using the human neuroblastoma cell line SH-SY5Y, expressing a range of voltage-gated sodium channel subtypes. SH-SY5Y cells were loaded with Calcium-4 No-Wash dye absorbed into the cells' cytoplasm. Cells were incubated for 5 min with CTXs prior to addition of veratridine. Fluorescence responses to CTXs, arising due to calcium ion influx into cells, were recorded using a Fluorescent Imaging Plate Reader. The performance of the SH-SY5Y assay was comparable to a N2a-based cytotoxicity assay [84]. The assay, however, is currently not widely used because it requires specialized equipment, the fluorescent dyes are expensive, and a small carryover of maitotoxins into the CTX fraction during purification steps could potentially obscure CTX responses due to saturation of fluorescence by maitotoxin-induced increase of intracellular calcium ions [3].

CTXs accumulate in fish tissue due to their lipophilic character, however, they also circulate in blood for some time. Bottein Dechraoui et al. were able to detect CTX in blood of mice after 12 h post-exposure of sublethal dose of Caribbean ciguatoxic extract (0.59 ng g−<sup>1</sup> C-CTX-1 equivalents), and pointed out that neuroblastoma assay (LOD 0.006 ng ml−<sup>1</sup> C-CTX-1) is suitable to monitor CTX in blood at sublethal doses in mice and argued that CTX monitoring in blood could be a useful procedure for fish screening [91]. Blood is a much simpler matrix than fish tissue, and CTX recovery from fresh blood is close to 100% due to lack of matrix effects [92]. Taking blood samples is a non-lethal sampling method for detection of CTXs in wild fish. O'Toole et al. studied the toxin level in tissue, blood and liver of the great barracuda (*Sphyraena barracuda*) [88] and observed a correlation between blood and liver toxin concentrations.

Although cytotoxicity assays are simple and provide an alternative to the MBA, they are currently not time- and cost-effective for mass screening, reflect only the combined potency of all-sodium-channel blocking toxins in an extract, and thus fail to provide any information about the toxin profile. Considering this latter point, however, limited information can be obtained by fractionating an extract using HPLC and then testing the toxicity of these fractions. Estevez et al. fractionated an amberjack sample, and CBA-N2a toxicity profile indicated the presence of at least four toxins (Figure 6b), which was confirmed by LC-MS/MS [51].

**Figure 6.** (**a**) Dose-response curve for N2a cells under OV<sup>+</sup> (cells treated with ouabain and veratridine mixture) conditions and under OV− (untreated cells without ouabain and veratridine added) conditions, exposed to increasing concentrations of four P-CTX standards, reproduced with permission from ref. [45]. Copyright MDPI AG, 2018; (**b**) CBA-N2a cytotoxicity profile of an HPLC fractionated sample, reproduced with permission from ref. [51]. Copyright MDPI AG, 2019; (**c**) Typical sigmoidal dose response curve of R-RBA. Bmax is the maximum binding of the bound radioligand (in counts per minute, CPM) in the absence of competing CTXs. Non-specific binding (NSB) represents the minimum total radioligand binding in the presence of saturating concentrations of CTXs, the EC50 is the effective concentration at 50%, reproduced with permission from [80]. Copyright Elsevier Ltd., 2018.


**Table 2.** Examples for the application of cytotoxicity and receptor-binding assays 1.

<sup>1</sup> CBA-N2a = mouse neuroblastoma cell assay; RBA = receptor binding assay (R-radioligand-based, F-fluorescence-based); n.a. = not available; <sup>2</sup> Composite toxicity (ng C-CTX-1 (TEF = 0.1), P-CTX-3C (TEF = 0.2) or P-CTX-1 toxin equivalent/g fish tissue); <sup>3</sup> Mass of fish tissue used for the analysis; <sup>4</sup> LOQ for CBA-N2a: (EC50 of P-CTX-3C/maximum concentration of dry extract)×(dry extract weight/fresh weight of flesh tissue extracted); <sup>5</sup> LOQ determination for RBA: LOQ <sup>=</sup> IC50 of P-CTX-3C/quantity maximum of a sample that does not cause matrix interferences; <sup>6</sup> Shark stomach sample; <sup>7</sup> Liver.

#### *4.5. Receptor-Binding Assays (RBA)*

CTXs compete with brevetoxin for the same neurotoxin receptor sites at sodium channels, having at least 20–50-times higher affinity [81], therefore CTXs are competitive inhibitors of brevetoxin binding [7,33,93,94]. Measuring the inhibition of radioactively labeled [3H]-brevetoxin-3 binding to rat brain membrane can be used to detect CTXs in fish extracts using either test tube [89,95] or microplate format [96]. Radioligand RBA (R-RBA) is based on competitive binding between CTXs and a radioactively labeled brevetoxin for a given number of available receptor sites in a membrane preparation. When the concentrations of the radioactively labeled brevetoxin and the receptor are kept constant and the concentration of CTXs increases, the binding of the radioactively labeled brevetoxin to the receptor sites is proportionally reduced. A competition dose-response curve can be constructed by measuring the concentration of the radioligand-receptor complex across a range of concentrations of CTX standard (Figure 6c), and the amount of CTXs in an unknown sample can be quantified [80]. RBA and CBA-N2a are more specific to CTXs than the MBA due to receptor binding, but RBA and CBA-N2a, like the MBA, do not provide any information on the toxin profile. The sensitivity of RBA is highly dependent on the receptor source. Bottein Dechraoui et al. compared the performance of the R-RBA and the CBA-N2a cytotoxicity assay, and the RBA was found to be 12-fold less sensitive than CBA-N2a for Caribbean ciguatoxin analysis [86]. It was also noted that R-RBA provided systematically higher estimates of CTX concentration than CBA-N2a [86,97]. Although R-RBA have been widely used, screening methods have not been standardized. Díaz-Asencio et al. recently published a methodological development and also determined criteria for quality control [80]. The developed microplate format was able to detect 0.75 ng g−<sup>1</sup> P-CTX-3C equivalent in fish tissue in only 3 h (process for full plate, not considering up to 2 d extraction time for 10 samples). An analyst could run an estimated 32 samples per day (with up to eight samples per plate run in triplicate at two dilutions) [80].

Large-scale field screening using R-RBA was performed by Darius et al. while evaluating toxic versus safe areas at two sites in French Polynesia, using the R-RBA test-tube format and rat brain synaptosomes as receptors [89]. Results indicated significant disagreement with the knowledge of local people regarding toxicity in several cases and findings were congruent only for fish species with high risk of CTX accumulation, indicating the need of close monitoring to prevent epidemiological outbreaks of CFP (see examples in Table 2, and more data in the original reference).

A disadvantage in the original radioligand format, and a constraint to application, is that R-RBA requires application of radioactive [3H]-brevetoxin compounds. To overcome this problem, McCall et al. recently developed a fluorescence-based RBA method (F-RBA) [98], which assesses competitive binding between CTXs and fluorescently labeled brevetoxin-2. Several fluorescent compounds, including BODIPY, 6-TAMRA, Texas Red, Alexa Fluor 350, Fluorescein, Coumarin, and Dansyl hydrazine, were conjugated to polyether brevetoxin-2, from which the BODIPY-brevetoxin-2 exhibited the best performance in terms of lowest nonspecific binding. The constructed F-RBA was faster (with an assay and analysis time of less than 3 h versus overnight), less expensive and safer than R-RBA, and generated binding constants comparable to the radioligand assay. As a continuation of this work, Hardison et al. developed a F-RBA using fluorescently labeled BODIPY®-brevetoxin-2 [97]. The method was relatively fast and took approximately 2 h to perform and exhibited a LOD and LOQ of 0.075 and 0.1 ng g−<sup>1</sup> P-CTX-3C equivalent, respectively. Based on this principle, a commercial test kit for detecting CTXs was developed by SeaTox Research Inc. (Wilmington, NC, USA; https://www.seatoxresearch.com/testing-kits/) and currently distributed by MARBIONC® Development Group LLC (Wilmington, NC, USA; http://www.marbionc.org/gallery/detail.aspx?id= 2274946). The SeaTox kit can be used as a screening or quantitative tool, and CTX analyses can be completed in less than two hours after fish tissue extraction.

#### *4.6. Immunoassays*

Immunoassays are dependent on the application of a high-affinity antibody that is selective and specific for CTXs, therefore, CTX detection and quantification with immunoassays are based on the structure of CTXs, not on their toxicity. In principle, a 100%-specific antibody could capture only one target, therefore detection of all toxins and toxin profile determination of a mixture of different toxins would require application of the same number of specific antibodies than the number of targets. If the antibody is less specific and selective only for a common structural motive of target molecules, it can capture several targets, but cross-reactivity with non-target compounds having the same structural element cannot be avoided. Production of antibodies is one of the key issues and constraints of immunoassays. In all cases, a label is attached to the antibody to detect target-antibody interaction; this label can be a radioisotope, enzyme, or fluorescent probe. Immunoassay methods, in general, are fast and easy to use, the sole exception being the radioimmunoassay (RIA). Unless several antibodies are used, they do not provide information on the toxin profile. Selecting antibodies only against the major toxin might also be problematic; P-CTX-1 in carrier fish, for example, represents in many cases greater than 90% of the toxins, but C-CTX-1 in fish represents only about 50% of total toxins [8]. Another problem that may rise with immunoassays is the possible cross-reaction of low-potency CTXs with antibodies, which increases the possibility of obtaining false-positive results. Antibodies developed for Pacific CTXs may not be suitable for testing Caribic CTXs, and *vice versa*.

The first immunoassay test, RIA, was developed by Hokama et al. in 1977 to assess fish tissues directly [99]. Labeled CTX antibodies were prepared by: (1) coupling purified CTX, isolated from toxic eel tissues, to human serum albumin; (2) injecting the CTX-human serum albumin conjugate into sheep for generating the anti-CTX-human serum albumin; (3) bleeding animals and collecting serum after 8 weeks; and (4) iodinating sheep anti-CTX-human serum albumin with 125I. This method was successfully used in the following years to test and reject toxic fish on the Hawaiian market [100], but it was time-consuming, expensive, and required special radioisotopic facilities, therefore it was impractical for routine screening of fish samples. The same sheep anti-CTX serum synthesized for

RIA was used by Hokama et al. to develop an enzyme-immunoassay (EIA) by coupling horseradish peroxidase (HRP) to sheep anti-CTX-human serum albumin instead of 125I [101]. 4-chloro-l-naphthol was used as substrate for the enzymatic reaction with spectrophotometric reading. EIA exhibited similar sensitivity and specificity to that of the RIA, yet, it was easier to run and economically feasible for screening fishes. However, this test remained labor intensive, and cross-reactions occurred with other polyether compounds. As a further development, Hokama constructed the first POC test utilizing the same sheep-anti-CTX antiserum [102,103]. The simplified enzyme immunoassay stick test (S-EIA) was rapid and did not require any instrumentation. The reagents for S-EIA were similar to those of EIA, and the salient feature of S-EIA was use of a coating (Liquid Paper, Liquid Paper Corp., Rockville, MD, USA) applied to a bamboo stick to adsorb the lipid CTX and its related polyether toxins onto the stick. Fish samples were poked with the stick and after fixation, the stick was immersed consecutively into sheep-anti-CTX-HRP solution and substrate solution. The substrate color change was read after 10 min, which provided a fast estimation of toxin content as being negative, borderline, or positive. All of these methods – RIA, EIA, and S-EIA—used fish flesh directly without extraction, but the antibody was not sufficiently selective, as the polyclonal sheep-anti-CTX detected not only CTXs but also structurally-related polyether toxins, including okadaic acid and brevetoxin [104,105]. This attribute resulted in false-positive tests [8,10]. Lack of sufficient sensitivity for S-EIA was also noted [106]. Due to these drawbacks, Hokama et al. prepared and used monoclonal antibodies, MAb, to CTXs in immunoassays [105]. Monoclonal antibodies are more selective to CTXs than are polyclonal variants, having some cross-reactivity with okadaic acid (16%), but little or none with polyethers (e.g., ionomycin). Using MAbs, Y. Hokama developed a simplified solid-phase colored latex immunobead assay (SPIA) for the field detection of CTXs and related polyethers (Figure 7) [107,108]. MAb-CTX was labeled with colored latex. In the simplified procedure, a paddle end of the stick coated with correction fluid (organic base solvent) is inserted into an incision in the fish so that the fish tissue is touched, then removed; the paddle end of the stick is then dried, fixed with methanol, and immersed into MAb-CTX-latex suspension. Visually detectable coloration of the stick is considered to be a positive test result. Hawaii Chemtect International (Pasadena, CA, USA) purchased the patents covering the S-EIA and SPIA tests and commercialized a rapid solid-phase immunobead assay (Ciguatect™) for the detection of ciguatera toxins [109]. The Ciguatect™ test kit could determine the ciguatera potential directly in the fish flesh or after toxin extraction. Toxin extraction increased the sensitivity and decreased the limit of detection [109]. The Ciguatect™ test was more sensitive than S-EIA and better suited to field applications [110]. (Note that the test is no longer used.) The S-EIA test required application of six assay sticks per fish [102,111].

Based on the same immunological principles as the SPIA, Hokama et al. developed a Membrane Immunobead Assay (MIA) using a polyvinylidene fluoride hydrophobic membrane laminated onto a solid plastic support to collect CTXs and a MAb to purified moray eel (*Muraenidae*) ciguatoxin attached to colored polystyrene beads [112]. Application of the hydrophobic membrane was an advantage in reducing non-specific binding of the immunobeads. The test procedure involved placing about 5 mg of fish tissue sample, 0.5 mL methanol, and the membrane stick into a test tube for 20 min, then removing, drying, and immersing this into 0.5 mL immunobead suspension (Figure 7). The intensity of the color on the membrane related to the concentration of the CTX. MIA exhibited much higher specificity than RIA, S-EIA, and SPIA [112]. Oceanit Test Systems, Inc. (Honolulu, HI, USA) marketed a POC test kit, Cigua-Check® test kit, what was based on MIA for P-CTX-1, and developed to test rice-grain-size amounts of fish flesh and to detect CTX higher than 0.05 ng g−<sup>1</sup> flesh [8]. In principle, this method was able to detect ciguatoxin at concentrations that induce clinical symptoms in humans (above 0.08 ng P-CTX-1/g fish flesh [8]). Concerns were, however, raised against the method's sensitivity and specificity, as well as the interpretation of the test-strip results [113], and possibly due to these and marketing issues, this method is currently not commercially available.

**Figure 7.** (**a**) Working principle of the simplified solid-phase colored latex immunobead assay (SPIA) test, adapted from ref. [107]; (**b**) Working principle of Membrane Immunobead Assay (MIA) test, adapted from [112].

Antibodies for antibody-based immunoassays are typically produced using scarce natural toxins. This limitation, in principle, could be overcome by replacing natural toxins with synthesized toxins, or with fragments of CTXs. The synthesis of fragments is more simple and cost-effective than that of the whole CTX molecule. These fragments, typically haptens, can be used to generate monoclonal antibodies against CTXs. Sandwich-type ELISA is very promising for increasing selectivity of assays by using two antibodies for the recognition of CTX, where one recognizes the left and the other the right wing of CTX. One of the MAb is conjugated with the enzyme label, typically HRP (Figure 8). To improve performance of immunoassays, Campora et al. constructed a sandwich-type ELISA, utilizing two antibodies, a chicken immunoglobulin Y specific to the ABCD domain of P-CTX-1 and a mouse monoclonal immunoglobulin G-HRP conjugate label specific to the JKLM domain of P-CTX-1 produced by injecting chicken and mouse with synthesized CTX fragments [114]. Significant cross reactivity was not observed for brevetoxin-3, okadaic acid, or domoic acid. Good correlation was observed between this ELISA and CBA-N2a assays by screening three fish species commonly implicated in ciguatera fish poisoning in Hawaii [115]. Tsumuraya et al. developed sandwich-type ELISA detection protocols for the four principal congeners of Pacific ciguatoxins, P-CTX-1, P-CTX-3, P-CTX-3C, and 51-hydroxy-P-CTX-3C, using MAbs produced by immunizing mice with the corresponding left and right wing haptens [116–119]. P-CTX-1 could be detected specifically as low as 0.28 ng mL−<sup>1</sup> without cross-reactivity with other related marine toxins; this concentration was still higher than the regulatory limit of 0.01 ppb [117]. As a further development, Tsumuraya et al. combined these MAbs into a single sandwich ELISA assay to detect any of these four CTX congeners [50,120]. Detection of the immunoreaction was changed to a fluorescent method using alkaline phosphatase(ALP)-linked MAb and a fluorescent substrate system, 2 -(2-benzothiazoyl)-6 -hyrodxybenzothiazole phosphate (BBTP). The fluorescent ELISA was highly sensitive, having a detection limit of less than 1 pg mL<sup>−</sup>1. P-CTX-1 spiked into fish at the recommended safety level of 0.01 ppb P-CTX-1 equivalent was reliably detected by this ELISA. The ELISA assay was shown to be very sensitive to CTXs, but required fish extract and laboratory conditions. Based on this ELISA, a fluorescent sandwich ELISA kit "CTX-ELISATM 1B" developed for detection of the P-CTX-1 series (P-CTX-1 and P-CTX-3) was commercialized and could be purchased from Fujifilm Wako Corporation (Osaka, Japan; cat. 382-14341) [120].

**Figure 8.** (**a**) Schematic of the sandwich ELISA detection of CTXs (*OPD* = *o*-phenylenediamine); (**b**) Illustration of the preparation of noncompetitive immunoassay with CE separation, adapted from [121].

Zhang et al. developed a capillary electrophoresis (CE)-based immunoassay for detection of P-CTX-1, applying electrochemical detection and gold nanoparticles (AuNPs) as carriers of HRP and CTX antibodies (Ab-AuNP-HRP) as target-capture elements [121]. Crude fish extract was sufficient for detection, which involved mixing the extract with Ab-AuNP-HRP probe solution, dilution, incubation, and analysis by CE separation and electrochemical detection. The unbound HRP, Ab-AuNP-HRP probe and the formed CTX-Ab-AuNP-HRP immunocomplex were separated according to the velocity difference in the separation capillary, and catalyzed the o-aminophenol (OAP) and hydrogen peroxide reaction to 2-aminophenoxazine-3-one (AP) (Figure 8). This latter was electrochemically detected. Due to the high separation power of CE and the high target specificity of the immunoassay, a LOD of 0.045 ng mL−<sup>1</sup> was achieved. As a further development, Zhang et al. fabricated an ultrasensitive immunoassay for P-CTX-3C detection exhibiting a LOD of 0.09 pg mL<sup>−</sup>1, based on CE separation and on-line sandwich immunoassay with rotating magnetic field [122]. The sandwich system utilized rabbit anti-P-CTX-3C-functionalized magnetite NP beads as immunosensing probes, and HRP and monoclonal sheep anti-P-CTX-3C-functionalized AuNPs as recognition elements. The rotating magnetic field enhanced the mixing efficiency and molecular binding rates, and the immunoreaction time of the assay was decreased to 15 min. The latter was much faster than normal ELISA.

The first electrochemical immunosensor for CTX detection was constructed by Leonardo et al. this year, 2020 [67]. Capture antibodies were prepared by immobilizing two different mouse MAbs on magnetic beads; these MAbs were able to bind to the left wing of P-CTX-1/P-CTX-3 and P-CTX-3C/51-hydroxy-P-CTX-3C, respectively. A mouse MAb, which was able to bind to the right wing of P-CTX-1, P-CTX-3, P-CTX-3C, or 51-hydroxy-P-CTX-3C, was biotinylated, linked to polyHRP-streptavidin signal reporter, and used as a detector antibody.

To develop the biosensor, the magnetic immunocomplexes were deposited on eight-electrode arrays, and the assay was performed by successively incubating the magnetic immunocomplexes with the CTX analyte and detector antibody. Although LOD and LOQ values of this CTX biosensor were higher than those of fluorescence ELISA, the electrochemical biosensor had an advantage of lower cost, the possibility to be integrated into compact analytical devices, and portability. The immunosensor was successfully applied to the analysis of fish samples and was able to detect P-CTX-1 at 0.01 ng g−<sup>1</sup> toxin level, as well as exhibited a good correlation with CTX levels determined by the CBA-N2a.

#### *4.7. HPLC, LC-MS*/*MS and LC-HRMS*

CFP is a complex disease in which several different compounds contribute to the toxicity of a fish sample. Current CBA-N2a, F-RBA, and immunoassays are very sensitive, but do not provide information about the toxin profile. Unfortunately, CTXs do not have characteristic functional groups for spectroscopic detection, therefore separating and separately identifying toxins is the only viable route for toxin profiles, and HPLC is the key method for separating CTX congeners. Classical HPLC methods using UV detectors, however, are not sensitive enough to detect very small concentrations of CTXs at clinically relevant levels, and CTXs cannot be distinguished using UV detection due to the lack of useful UV chromophores [12,33,123,124]. At a terminus of the molecule, however, most of the CTX congeners have a primary hydroxyl group (see Figures 1 and 2) that is available for fluorescence labelling. HPLC with fluorescence detection using 1-anthrylcarbocyanide and carbonyl azides or carbonyl nitriles of coumarin derivatives as labels was found to be more sensitive for CTX detection than HPLC-UV, however, still could not reach the recommended tolerance level of 0.01 ng g−<sup>1</sup> for P-CTX-1 [12,125,126]. The main limitation of this technique was that it could not detect CTXs without a primary hydroxyl group (e.g., P-CTX-3C; see Figure 1), what is especially problematic for herbivores where CTXs without primary hydroxyl group are more abundant [127].

To utilize the separating power of HPLC, a sensitive detector is required. HPLC coupled with tandem mass spectrometry (HPLC-MS/MS) were applied for CTX detection the first time by Lewis and Jones in 1994 [128], and the first successful application of the combined HPLC-MS/MS technique for CTX detection at clinically relevant levels, with detection limits of at least 0.04 ppb for P-CTX-1 and 0.1 ppb for C-CTX-1, was reported by Lewis et al. in 1999 using gradient reversed-phase HPLC and an electrospray triple quadrupole mass spectrometer [129]. The method was significantly more sensitive than fluorometric HPLC. In the following decades LC-MS/MS became one of the leading and indispensable techniques of modern CFP laboratories due to the method's ability to separate and identify toxins, and its uniqueness in providing toxin profiles. Both the identification and quantitation of CTXs, however, require reference toxin molecules. Although the molecular peak and fragmentation pattern in MS can provide valuable information for CTX identification, a diagnostic fragmentation pattern is not always achieved (water losses are often the most abundant fragment ions and depending on the MS instrument used, results may be different). LC-MS/MS must be used in combination to biological assay as a confirmation. In general, raw fish extract could be used for LC-MS/MS, but to decrease possible interference of lipids and fatty acids in fish tissue, the raw extract is usually purified using SPE in one or two steps (Figure 3). The presence of matrix co-extractives significantly interferes with ionization and causes severe signal suppression [49]. To reach very low detection limits, both HPLC and MS conditions must be optimized for sensitivity and selectivity, namely: LC conditions, ionization sources, ion choices and acquisition modes [41,46,130,131]; a comparison of analytical protocols to find the best conditions for sensitivity and/or selectivity for LC-MS/MS has been published recently [130].

LC separation is most frequently performed with acetonitrile-water or methanol-water linear gradients; the methanol-based mobile phases have been found to be more advantageous. Electrospray ionization and positive ion detection mode, using either Single Ion Monitoring (SIM), Multiple Reaction Monitoring (MRM) or Enhanced Product Ion (EPI) scanning, are typically used in MS. Applying MRM is advantageous for increasing specificity. The mobile phase composition, solvents and additives, and flow rate of the mobile phase affect ionization efficiency and ion abundance, therefore they have to be optimized. Monitoring ammonium or sodium adduct as the parent ion often exhibited higher signal-to-noise ratios than protonated parent ion; however, this depends not only on LC and MS conditions, but also on the analyzed CTX (Figure 9a). With formic acid, as a mobile phase additive, the sodium adduct was favored using either acetonitrile/water or methanol/water linear gradients in LC. The sodium adduct, however, has been found to be relatively stable, therefore to obtain MS/MS fragmentation its generation should be avoided by proper selection of eluent and additives, e.g., ammonium formate [130]. The ammoniated adduct ion, as a precursor ion, showed an

advantage for selectivity through confirmatory transitions [130]. Figure 9b shows an example for the LC separation and detection of a mixture of CTX standards. The technique using triple quadrupole detectors (low resolution, LC-MS/MS) is significantly more sensitive than high-resolution MS coupled to HPLC (LC-HRMS, time-of-flight and orbitrap based spectrometers). An important advantage of HRMS, however, is that HRMS provides molecular formula and isotopic patterns of molecules (Figure 9c), therefore providing improved identification of CTX analogues [130,132]. LC-HRMS together with reference material may be the best choice for accurate identification of CTXs. The two techniques may be considered complementary, as LC-MS/MS is more adequate for quantitation and LC-HRMS performs better for identification of toxins. Selected examples for the application of LC-MS for CTX quantification in fish tissue, including fish species, toxin content and LOQ, are shown in Table 3 (more data are provided in the cited articles). Fish blood for determining the CTX content was studied recently by Mak et al. [133], and the matrix effect was found to be smaller than that of using fish muscle tissue. The main advantage of LC-MS is that it is very specific and adequately sensitive to detect CTXs at the clinically relevant toxin levels, and superior to any other techniques as the toxin confirmatory method.

**Figure 9.** (**a**) LC-MS/MS chromatogram of standard CTX solutions showing that the [M + Na]<sup>+</sup> and [M + NH4] <sup>+</sup> ions were favored for P-CTX1B (P-CTX-1), while the [M + NH4] <sup>+</sup> and [M + H]<sup>+</sup> ions were dominant with similar intensities for P-CTX3C (P-CTX-3C) at identical conditions. A linear gradient of water-methanol solvent mixture, containing ammonium formate and formic acid, was used for chromatographic separation. SIM was performed in positive mode for ions at *m*/*z* [M + Na]+, [M + H]<sup>+</sup> and [M + NH4] <sup>+</sup>. Reproduced with permission from ref. [130]; (**b**) LC-MS/MS chromatogram and retention time of P-CTXs standards using a Zorbax C18 column, water and methanol as eluents, and a linear gradient. The separation of various CTX congeners is shown. Reproduced with permission from [130]. Copyright Elsevier Ltd., 2018; (**c**) LC-HRMS high resolution mass spectra of CTX1B (P-CTX-1) obtained from CTX standard, showing adduct peaks and isotopic patterns, reproduced with permission from [132]. Copyright Elsevier Ltd., 2017.


**Table 3.** Examples of the application of LC-MS for CTX detection.

<sup>1</sup> Mass of fish tissue used for the analysis; <sup>2</sup> Concentration of toxin (ng toxin/g fish tissue); n.a. <sup>=</sup> not available; <sup>3</sup> Due to the relationship between the limit of detection, LOD (S/<sup>N</sup> <sup>&</sup>gt; 3), and limit of quantitation, LOQ (S/<sup>N</sup> <sup>&</sup>gt; 10), only LOQ is shown in table (ng toxin/g fish tissue); <sup>4</sup> total toxicity; <sup>5</sup> Shark stomach; <sup>6</sup> in P-CTX-1 equivalent.

#### **5. Outlook and Conclusions**

CFP is an old problem for communities in tropical regions that rely on seafood for survival, and CFP has now become a global issue due to the worldwide seafood trade, international travel, and ongoing expansion of the geographic ranges of fish contaminated with ciguatoxins. There is, therefore, high demand to test fish for CTXs prior to human consumption, both for mitigation of the health risks of CFP and for clinical identification of toxins in cases of intoxication. Currently, however, reliable biomarkers that can confirm exposure to CTXs and accepted diagnostic tests for direct detection of CTXs in humans are not available; therefore, diagnosis of ciguatera relies on clinical observations of its overt symptoms and/or testing remnants of consumed fish, if available. Several methods have been developed to screen for the presence of CTXs in fish tissue prior to consumption. In vivo whole-animal detection methods are now superseded by in vitro assays that have greater sensitivity, including receptor-binding assays (RBAs), cell-based assays (CBAs), Enzyme-Linked Immunosorbent Assays (ELISA), capillary electrophoresis (CE)-based immunoassays, electrochemical immunosensors (ECS), and liquid chromatography tandem mass spectrometry (LC-MS/MS). Currently employed methods are compared in Table 4. Present methods for CTX analysis, in general, are labor-intensive, time-consuming, and require laboratory facilities with well-trained technicians. To date, these methods have not been properly validated. At present it is difficult to obtain sufficient standard CTXs as reference calibrants, impeding corroboration and widespread application of these analytical techniques.


**Table 4.** Comparison of methods for CTX detection.

<sup>1</sup> Depends on purity requirements. Parallel samples can be prepared simultaneously depending on lab conditions and operator. Extraction time is different depending on the protocol used. See also Figure 3. As an example, the estimated time for preparing the raw extract is 5–6 h, and for the two SPE purification step is 2 h; <sup>2</sup> 96-well microtiter plate is typically used for RBA, CBA and ELISA; however, sample throughput can be 96–1436; <sup>3</sup> LOQ for P-CTX-1 equivalent toxin in fish tissue (protocol dependent), note that the suggested tolerance limit for P-CTX-1 in fish tissue is 0.01 ng g−1, respectively; <sup>4</sup> Includes 24-h incubation and 24-h exposure of the neuro-2a cells to fish extracts, and 4–6 h assay time; extract preparation can be undertaken during incubation time; <sup>5</sup> LOD, no specific measurement for LOQ was done; <sup>6</sup> This estimation considers only the time of a single injection of a sample into the LC-MS system, and does not includes calibration and quality controls.

Detection of CTXs in fish tissue is not simple, due to their generally low concentrations, the presence of multiple CTX congeners, the limited amount or lack of CTX reference materials for many derivatives, the difficult synthesis or lack of CTX antibodies, the co-occurrence of interfering molecules in fish tissue, and the typically unpredictable incidence of CTXs. Not all of the currently used methods for quantification of CTXs offer sufficient sensitivity, specificity and selectivity. The mouse bioassay (MBA) and radioligand receptor-binding assays (R-RBA) are expected in the near future to be replaced in laboratories due to their low sensitivity, lack of specificity, and high cost, as well as ethical and safety concerns. The fluorescence-based receptor-binding assay (F-RBA) eliminates safety issues associated with radioactive compounds, and its wider application is expected in the future, especially due to the commercial availability of SeaTox® F-RBA test kit. One of the most successful screening methods seems to be the cell-based assay (CBA) using mouse neuroblastoma cells (N2a). This method is simple and sufficiently sensitive, but is time consuming and reflects only the combined effect of the various CTXs present in fish extract. Further, other toxins can also block sodium channels, therefore, the method fails to provide information on toxin profiles. Immunosensors and immunoassays in the form of ELISA or CE-based tests are sensitive and selective tools for ciguatoxin detection, however, they are limited to the availability of CTX-antibodies. Although, in principle, several toxins can be targeted simultaneously, assays typically contain one or a very limited number of antibodies, therefore these methods do not provide information about toxin profiles. Detecting specific toxins can be misleading because toxin profiles can vary considerably among fish. Interestingly, electrochemical biosensors, except for CE and the recently constructed electrode array ECS, have not been developed for CTX detection in fish tissue, possibly due to the lack of sufficient amounts and types of CTX antibodies. Solving the antibody problem in the future may stimulate large-scale development in this field, especially since antibody-based methods offer the potential for miniaturization and development of portable devices and POC tests [139–141]. The application of LC-MS/MS for CTXs detection was an important break-through in CTX screening, because this method is sufficiently sensitive, selective, and unique in providing toxin profiles. The major obstacle of its wide-scale application is the lack of widely available reference toxins for quantification. Currently, state-of-the-art CTX detection involves a combination of CBA-N2a assay and LC-MS/MS, where CBA-N2a assay is used for screening the total toxicity of the sample, and LC-MS/MS to confirm the toxins and for providing toxin profiles [19,45,51,87]. LC-MS/MS together with reference materials can fulfill the role of the sensitive and specific CTX detection method, however, it requires highly trained operator and laboratory facilities. Further the method cannot be miniaturized and field application is not possible.

Testing fish tissue for CTXs is of crucial importance because typically muscle tissue is consumed by humans. Such testing, however, requires a highly invasive sampling method, therefore it is not suitable for testing protected fish species, environmental risk assessment, or clinical diagnosis of human ciguatera fish poisoning. Collecting blood samples is not particularly invasive, and has an added advantage that blood is a much less complicated matrix for analysis than muscle tissue. Unfortunately, few studies have been conducted for testing CTXs in fish blood samples, and the possible relationship between the concentrations of CTXs in muscle tissue and blood, if one indeed exists, is not yet known—elucidation of the functional association between these two parameters though statistical modelling would be highly beneficial towards effective screening of the toxin. It is, however, expected that circulating CTX content in blood is higher after a meal of ciguateric fish. Future research will likely focus in this area because it is especially important for identifying CTXs in humans following its intoxication.

Presently, there is great demand for a portable, fast, reliable, easy-to-use, and cheap CTX screening assay for private customers, fishermen, and fish vendors to test their food or catch before consuming or selling it. These point-of-care tests, however, currently do not exist, notwithstanding that performance requirements may be lower than for those methods used in laboratories. CTX testing is still in the hand of specialists, and specialized laboratories are now able to provide sufficiently accurate information about toxin profile and toxin concentration in fish. Although significant advances have been made to develop and improve the performance of ciguatera assays, the ideal assay, one that is simple, rapid, reliable, robust, highly sensitive, quantitative, provides specific toxin profiles, cheap and does not require trained operators and specialized equipment, does not exist to date; developing one is the key challenge for future research on this field.

**Funding:** This research received no external funding.

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

#### **References**


© 2020 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 (http://creativecommons.org/licenses/by/4.0/).

## *Review* **Ciguatera in the Indian Ocean with Special Insights on the Arabian Sea and Adjacent Gulf and Seas: A Review**

**Nazima Habibi 1, Saif Uddin 1,\*, Marie-Yasmine Dechraoui Bottein <sup>2</sup> and Mohd Faizuddin <sup>3</sup>**


**Abstract:** The dinoflagellates of the genus *Gambierdiscus* are found in almost all oceans and seas between the coordinates 35◦ N and 35◦ S. *Gambierdiscus* and *Fukuyoa* are producers of ciguatoxins (CTXs), which are known to cause foodborne disease associated with contaminated seafood. The occurrence and effects of CTXs are well described in the Pacific and the Caribbean. However, historically, their properties and presence have been poorly documented in the Indian Ocean (including the Bay of Bengal, Andaman Sea, and the Gulf). A higher occurrence of these microorganisms will proportionately increase the likelihood of CTXs entering the food chain, posing a severe threat to human seafood consumers. Therefore, comprehensive research strategies are critically important for developing effective monitoring and risk assessments of this emerging threat in the Indian Ocean. This review presents the available literature on ciguatera occurrence in the region and its adjacent marginal waters: aiming to identify the data gaps and vectors.

**Keywords:** the Indian Ocean; Arabian sea; Kuwait bay; Aden Gulf; Red Sea; Gulf of Aqaba; Andaman Sea; Bay of Bengal; seafood safety; foodborne disease

**Key Contribution:** This review highlights the paucity of data on ciguatoxins from the vast Indian Ocean region, which is home to an enormous mass of seafood-consuming populations. Furthermore, it exports fisheries to Europe and North America, which resulted in toxic fish being found during 2012, 2015, and 2016. The monitoring of *Gambierdiscus* and *Fukuyoa* is not taken up regularly despite their recorded presence all over the Indian Ocean.

### **1. Introduction**

Ciguatera poisoning (CP) is a syndrome caused by ingestion of coral reef fish and shellfish of tropical and subtropical regions, which has caused global concern. Some dinoflagellate species of the genera *Gambierdiscus* and *Fukuyoa* are known to produce ciguatoxins (CTXs); the organisms that consume these toxic algae accumulate CTXs that are transferred and biotransformed along the marine and human food chain. These lipidsoluble and heat-resistant toxins cause gastrointestinal, cardiovascular, and neurological disorders among humans consuming the CTX-contaminated seafood [1–5]. According to a study published in 2008, CP disease estimations are uncertain and often misdiagnosed [6]. However, few reports predict 2–10 million people are annually affected by CP [7,8].

Ciguatera poisoning is regarded globally as the most significant non-bacterial poisoning associated with fish consumption. It is usually limited to the consumption of toxic fish from regions between the latitudes 35◦ N and 35◦ S [9–17]. Studies have shown a strong positive correlation between *Gambierdiscus* abundance and algal macrophytes [18]. Some earlier studies proposed a standardized methodology for estimating *Gambierdiscus* abundance based on sampling macroalgae [19–21]. More recent studies brought to light the significant biases in *Gambierdiscus* cell distribution within the macrophytes. A 33–150%

**Citation:** Habibi, N.; Uddin, S.; Bottein, M.-Y.D.; Faizuddin, M. Ciguatera in the Indian Ocean with Special Insights on the Arabian Sea and Adjacent Gulf and Seas: A Review. *Toxins* **2021**, *13*, 525. https://doi.org/10.3390/ toxins13080525

Received: 2 July 2021 Accepted: 22 July 2021 Published: 27 July 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/).

variation was reported among replicates [22–24]. Despite these limitations, some first-order estimates on *Gambierdiscus* distributions for large geographic regions have been attempted using the average abundances [11,25]. These have shown a global distribution spreading across oceans with 85% of *Gambierdiscus* density estimates <1000 cells g−<sup>1</sup> wet weight algae, with <10% occurrence with 10–10,000 cells g−<sup>1</sup> wet weight algae and <5% incidences where the concentration was >100,000 cells g−<sup>1</sup> wet weight algae. With the growing global demands for seafood and international trade ease, CP concerns the population beyond the endemic regions and is increasingly becoming a global issue. Moreover, a greater threat is posed as there are no current antidotes for CP [26].

Historically, poisoning associated with seafood consumption was reported in different parts of the globe. It was first recounted in the West Indies as early as 1511 [27] and in the Gulf of Guinea in 1521, killing the Captains of the Spanish army [28]. It was also reported in the islands of the Indian Ocean in 1601 and various archipelagos of the Pacific Ocean in 1606 [27]. Additionally, in 1786, the surgeon of HMS Endeavour, en route to Australia, New Zealand, and the Pacific Islands, reported that the ship's Captain was poisoned by ciguatoxins [29]. A year later, the ingestion of a local gastropod (*Livonia* sp.) was discovered to induce neurological symptoms in the Antilles of the Caribbean Sea [30]. In 1948, the organism *Gambierdiscus* (originally referred to as *Goniodoma* sp.) was described for the first time in Cabo Verde [31]. The term "ciguatera poisoning" was coined in Cuba (Caribbean Sea) after the ingestion of a marine snail (*Turbo pica*) locally known as cigua [32]. The historical facts suggested the tropical and subtropical Pacific and Indian Ocean insular regions, the tropical Caribbean region, and the continental reefs to be endemic to ciguatera [27].

In recent years, there has been an increase in the frequency of reports of toxic and harmful benthic algal blooms, predominantly *Gambierdiscus* with the presence of *Fukuyoa* and *Ostreopsis* blooms, throughout the world [33,34]. The interest in *Gambierdiscus* blooms has been heightened because of increased awareness of the effects of CP on human health. A growth in species abundance has already been observed in subtropical and temperate regions with the threat of global warming expected to further exacerbate the situation [35–38]. However, a recent article by Hallegraeff et al. [39] suggested that intensified monitoring efforts and heightened aquaculture activities are responsible for these perceived increases in HABs events that are not underpinned to be expanded as an empirical assumption. Ciguatoxins are reported globally, being described from new sites in the Canary Islands, Indian Ocean, Japan, and Western Gulf of Mexico [8,40–45]. Until recently, the records of *Gambierdiscus* in the Indian Ocean were scarce and restricted to the western tropical region, whereas now its presence has been found in the Northern part of the ocean [42]. There is paucity of information on ciguatera phenomena, including the occurrence of human poisonings, of toxins in seafood, and of the causative organisms, in the Indian Ocean in general, and the northern part in particular. In the present review, we gathered the published information on reported occurrences of *Gambierdiscus* and identified the research gaps related to its monitoring as a tool to manage this emerging hazard.

#### **2. Results**

#### *2.1. Environmental and Global Pressures in the Northern Part of the Indian Ocean*

The Indian Ocean is the third largest ocean enclosing densely populated landmasses: In the North, India, Bangladesh, Burma, Thailand, Pakistan, Iran, and Oman among others. It also includes regional seas in the North such as the Arabian Sea and the adjacent Lakshadweep Sea, Aden Gulf in the Red Sea, the Gulf of Aqaba and Suez, the Bay of Bengal and the Andaman Sea, as well as the Gulf of Oman and the Persian Gulf further north. Countries from the Indian Ocean region have experienced unusual climatological conditions such as cyclones, El Niño-Southern Oscillation events, and coral reef bleaching during the past two decades [46–48]. Benthic microalgae are particularly influenced by these disturbances especially the coral mortality, which provides a good substrate for the formation of algal turfs and associated epiphytes [49]. The study conducted by

Quod et al. [46] reported a significant increase in the CP causative organisms in comparison to those reported in 1980 [47], following the coral bleaching event in 1998, thus posing an incremental risk of HABs in the region and drawing detrimental consequences to the marine biodiversity and human health [50].

The unfavorable effects of increasing atmospheric levels of carbon dioxide (CO2) and other greenhouse gases in the Indian Ocean and its marginal seas are leading towards acidification of the marine environment [48,51–53]. Supporting evidence for increased CO2 sequestration was drawn from increased marine primary productivity over the past decade [51,54]. The eutrophication of Kuwait bay and the nearby water bodies due to upwelling events is considered to be a predominant factor influencing the onset of algal blooms due to enriched nutrient conditions [54]. Other potential factors that have influenced the onset of blooms consist of coral bleaching, unusual variances in temperature, and calm conditions. Additional factors such as dust storms that carry micronutrients, domestic and industrial inputs, natural meteorological and oceanographic forcings, and the introduction of invasive species from ballast water discharge may all play a major role in the onset and expansion of HABs [41,45,55–58].

#### *2.2. Ciguatera Causative Organisms Occurrence*

The dinoflagellates *Gambierdiscus* and *Fukuyoa* are the causative organisms of CP worldwide [5]. The genus *Gambierdiscus* has a pantropic distribution with about 18 known species while *Fukuyoa* has 3 known species. Within the Indian Ocean, *Gambierdiscus* are more dominant in the western part. Of the several known *Gambierdiscus* sp., initially only *Gambierdiscus toxicus* was reported in Mayotte since *G. toxicus* was the first species that was described [50,59–61]. Later it was reported in La Reunion and Mauritius [62–65] as well (Figure 1). *Gambierdiscus toxicus* was also found in Mbudya Island, Oysterbay, Bawe Island, and Makoba Islands in Tanzania at depths of 5–10 m and temperatures ranging from 25 to 32 ◦C [66] (Figure 2). Thereafter, *Gambierdiscus yasumotoi* (now *Fukuyoa yasumotoi* (M.J. Holmes) [67]) and *Gambierdiscus belizeanus* were confirmed to occur in Mayotte in the Comoros archipelago [50] (Figure 3).

**Figure 1.** Spatial distribution of *Gambierdiscus* and *Fukuyoa* sp. from Reunion Island.

**Figure 2.** The occurrence of *Gambierdiscus* sp. from Tanzanian coastal area.

**Figure 3.** *Gambierdiscus* sp. reported from Mayotte Island, Indian Ocean.

In the northern part of the Indian Ocean, *Gambierdiscus* species have been noticed in the marginal seas of the Indian Ocean, such as the Arabian Sea [41,42,45,68]; Red Sea [41–45]; Gulf of Aqaba [42]; and Manora Channel, Pakistan [69]. An assessment carried out in Kuwait (Persian Gulf) from November 2012 to March 2013 showed the presence of *G. yasumotoi* in the shallow lagoons of Qit'at Julai'ah and southern coastal waters of Kuwait at 1–3 m depth [68,70] (Figure 4). The average seawater temperatures and salinity during this time were between 23.5–25 ◦C and 41.2–42.4 ppt, respectively. In another study, *G. toxicus* was also identified in Kuwait's territorial water [41,45] (Figure 4).

**Figure 4.** *Gambierdiscus toxicus* and *Fukuyoa yasumotoi* occurances in Kuwait Coastal waters.

In an investigation from the Gulf of Aqaba, Jordan during October 2010, with a seawater temperature range of 24–25 ◦C and salinity of 40 ppt, *G. belizeanus* was reported from a depth of 1.5–2.0 m (Figure 5). The toxigenic *G. belizeanus* was reported from the Red Sea, Saudi Arabian coast during February 2012 and May 2013 [43,44] at a sampling depth between 0.4–0.8 m, with seawater temperatures and salinity range of 28.9–38.4 ◦C and 36.1–38.4 ppt respectively. The oxygen saturation during this period was very variable between 1.73–32.1%.

**Figure 5.** The Spatial distribution of *Gambierdiscus belizeanus* and *Fukuyoa yasumotoi* occurances in Red Sea and adjacent areas.

In 2011, four *Gambierdiscus* species including *G. toxicus*, *G. belizeanus*, *G. polynesiensis*, *G. australes*, and *Fukuyoa yasumotoi* were also reported from the coastal waters of Pakistan in the northern Indian Ocean [69] (Figure 6). An unidentified *Gambierdiscus* sp. was also discovered in middle of the Bay of Bengal [71] (Figure 7). In Mangalore at the southwest coast of India, CP was first described during June 2015 outbreak involving red snapper(*Lutjanus bohar*) [72,73], and a second outbreak during September 2016 involving red snapper again [74]. In another incidence, CP was also reported from Trivandrum, India [75].

**Figure 6.** Occurrences of *Gambierdiscus* in Gulf of Oman and Pakistan's coastal regions.

**Figure 7.** *Gambierdiscus* sp. reported from the Bay of Bengal, Indian Ocean.

#### **3. Discussion**

#### *3.1. Toxin Production by Ciguatera in Indian Ocean Region and Adjoining Marginal Seas*

Ciguatoxins form a large group of toxins of 40 confirmed or suspected chemical analogs. They are classified into four types, according to both their geographical origin and chemical structure: The Pacific CTXs (P-CTX) of the type CTX3C and CTX4A, the Caribbean CTXs (C-CTX type), and the Indian CTXs (I-CTX type) [8,27]. The existence of the I-CTX type is still speculative and may be similar to that of the C-CTX. Their structural characteristic has not yet been elucidated. Six derivatives I-CTX 1–6 were identified with a molecular weight in Dalton of 1140.6, 1156.6, 1138.6, and 1154.6, however, their chemical structure is still unknown. Four of them were retrieved from a highly toxic *Lutjanus sebae* (Red Emperor) from the coast of the Republic of Mauritius through optimized gradient reversed-phase high-performance liquid chromatography-electrospray ionization mass spectrometry (LC/MS) methods, in combination with a radioligand receptor binding assay (r-RBA) [76,77], and two from in a bull shark (*Carcharhinus leucas*) implied in a fatal intoxication in Madagascar [76–78]. The lipid-soluble extracts retrieved from the edible fishes were reported to possess CTX activity and induced lethal symptoms in mice [77] and responded to the cell bioassay [78].

The toxicity of these *Gambierdiscus* species collected in the Pacific and/or the Caribbean was found to vary by over 2 orders of magnitude [11,79,80]. It remains to be confirmed that the same ranges are observed in the Indian ocean where only a few species have been analyzed for their toxin contents. Using in vitro cytotoxicity cell-based assays (CBA with neuroblastoma with N2a cells), the toxicity of two isolates of *G. belizeanus* from the Red Sea was estimated at 6.50 × <sup>10</sup>−<sup>5</sup> pg P-CTX-1 eq. cell−<sup>1</sup> for RS2-B6 and 1.02 × <sup>10</sup>−<sup>5</sup> pg P-CTX-1 eq. cell−<sup>1</sup> for RS3-B8. Toxin production was slightly higher in *G. belizeanus* strains from Tahiti (0.0246 fg P-CTX-3C eq. cell−1) [4,43,44]. As compared to *G. belizeanus,* the species of *G. polynesiensis*—0.017–4.4 pg P-CTX-3C eq. cell−1; *G. toxicus*—0.028 pg P-CTX-3C eq. cell−<sup>1</sup> [4]; and *G. australes*—0.04 pg CTX-1 eq cell−<sup>1</sup> [40] are more toxigenic.

More recently, Gambieric acid D was identified for the first time in the flesh of a bull shark (*Carcharhinus leucas*) employing the technique of liquid chromatography coupled with high-resolution mass spectrometry (HRMS). The toxicity was confirmed through mouse bioassays (Lowest dose = 72 mg equiv. stomach per mouse of 20–22 g) and neuro-2a cellbased assays (flesh—0.06 μg P-CTX-1 equiv./kg; stomach—92.78 μg P-CTX-1 equiv./kg; fins 1—0.12 μg P-CTX-1 equiv./kg; and fin 3—μg P-CTX-1 equiv./kg) [78]. Ciguatoxins were identified as CTX1B and 2,3-dihydroxy CTX3C; 51-hydroxy CTX3C; or positive by cell assay in Snapper fish (Lutjanidae) caught in the Indian Ocean and involved in 5 CP outbreak [81]. Fish samples from Sri Lanka were analyzed by the European Union Reference Laboratory for Marine Biotoxins (EURLMB) in Vigo, Spain, and the putatively contaminated fish samples were analyzed by liquid chromatography-tandem mass spectrometry (LC-MS/MS) according to a method published by [82] with slight modifications. The only available analytical standard was P-CTX-1B. Seven out of 11 samples tested positive for P-CTX-1B. In addition, other putative CTX variants were detected across most samples, however, they could not be confirmed as CTXs due to the lack of reference compounds. There are incidences of ciguatera poisoning due to the import of contaminated fish from the Indian ocean region, mainly India and Sri Lanka.

It is important to note that the investigations conducted so far have been limited in spatial coverage. Additional research is, therefore, required at a broader spatial scale to make meaningful conclusions regarding the occurrence, abundance, and temporal variability of ciguatoxins in the region.

#### *3.2. Ciguatera Causative Organism's Abundance*

Globally, ciguatera causative organisms' densities have been found to vary from few cells to thousands of cells per gr of substrate. *Gambierdiscus* is known as a slow-growing species compared to many other dinoflagellates; it takes about 5 months to increase its abundance. Chinain et al., [4] found a mean growth rate in *G. polynesiensis* as 0.13 ± 0.03 div d−<sup>1</sup> at exponential phase in batch culture condition [4]. In the field, a 17 months lag time is estimated between a major environmental event (such as hurricane) and before it blooms and releases toxins into the environment and gets into the food chain [83]. During October 1998, a remarkably higher concentration of *G. toxicus* equivalent to 60,463 cells per gram of algae was

recorded in Mayotte Islands (Comoros, southwest Indian Ocean) following a bloom event. This density is the highest ever recorded in the region and also the highest globally [50]. In a non-bloom event, the cell densities of *Gambierdiscus* were far less and often non-detectable.

In the western Indian Ocean region, the cell densities of *G. toxicus* were variable and affected by spatial and temporal variations. In Tanzania, it ranged from 0.0 to 879.5 (Bawe station) cells g−<sup>1</sup> wet weight (ww) algae and 0 to 92.6 cells g−<sup>1</sup> ww seagrass (Mbudya) [66]. The abundances on macroalgae in the Mayotte coral reef complex were reported to range from 0.0 to 2800 cells g−<sup>1</sup> [60]. Monitoring of *G. toxicus* in the locality of Saint Leu (Reunion) since 1993 revealed high variability in population density with an average value of 122 ± 24 cells g−<sup>1</sup> of algal turf [64]. In the coral reef complex of Mayotte Island (SW Indian ocean), the concentrations of *G. toxicus* ranged from 800–5400 dinoflagellates g−<sup>1</sup> of biodeposits in the northeast lagoon. Up to 6000 dinoflagellates g−<sup>1</sup> were recorded at Bambo islet in the southeast lagoon. Lower abundances up to 400 g−<sup>1</sup> were recorded near the main island shore; seawards on the outer side of the barrier reef; and in luxurious coral areas more exposed to humans. At stations near the main island coast, abundance of *G. toxicus* was less than 100 cells g−<sup>1</sup> [60]. The density of *G. toxicus* was as low as 0–4 cells g−<sup>1</sup> of macrolagae in the lagoons of Trou Aux Biches, Mauritius [84].

In the northern part of the Indian ocean, the cell numbers of *Gambierdiscus*from the central Red Sea, Saudi Arabian coast were <40 cells g−<sup>1</sup> ww of algae [43,44]. Similarly, *F. yasumotoi* in Kuwaiti shores had an average cell density of 116.7 ± 47.5 cells g−<sup>1</sup> of ww algae. The high biomass algae (HABs) were estimated through the Ocean Color Modis Algorithm (OC3M), Garver-Siegel-Maritorena Algorithm (GSM), Generalized Inherent Optical Property (GIOP) model [41,45]; these areas when sampled showed the presence of *G. toxicus* in a concentration of ~1000 cells per liter of seawater [41].

#### *3.3. Ciguatera Causative Organisms' Substrates and Co-Occurring Species*

*Gambierdiscus* is an epiphytic benthic dinoflagellate, commonly found on algal turfs, coral rubble, and macroalgae. A study from the Coastal region of Tanzania reported the *G. toxicus* strains along with unknown brown algae mixed with filamentous cyanobacteria species and the red alga *Gracillaris* sps. [66]. The dinoflagellate was also separated from two species of seagrass namely *Thalassia hemprichii* and *Thalasodendron ciliatum* in the same study [66]. The dinoflagellate assemblage observed in coral reefs revealed toxic species including *G. toxicus*, *Prorocentrum* spp., and *Ostreopsis* spp [85]. The co-occurrence of *Prorocentrum* sps. and *Ostreospis* sps. with *G. toxicus* were reported by Grzebyk and his group. In the same study, it was also demonstrated that some microalga stimulated whereas the others subsided the growth of *G. toxicus* [60]. Benthic thecate dinoflagellates in the sandy ecosystem of La Possession bay (Reunion Islands) were isolated as a complex of five toxic species namely *G. yasumotoi*, *G. toxicus*, *P. arenarium*, *P. concavum,* and *P. lima*. At St. Leu and Saline reef, 25 benthic thecate dinoflagellate species coexisted out of which 15 were harmful and are presented in Table 1 [65]. The benthic dinoflagellate of *G. toxicus* was reported to coexist with *Ostreopsis* spp., *Prorocentrum* spp., *Coolia monotis,* and *Amphidinium* sp. in the lagoon of Trou Aux Biches, Mauritius [84]. In Mayotte islands, South west of Indian Ocean, *G. toxicus* were reported along with *Prorocentrum* spp., *Ostreopsis* sp., and *Amphidinium* spp. [50,59].


**Table 1.** Occurrence of *Gambierdiscus* species in the Indian Ocean region.


**Table 1.** *Cont.*


**Table 1.** *Cont.*


**Table 1.** *Cont.*

NA-Not available.

In the reports of Catania and her team [43,44], the Red Sea strains of *G. belizeanus* were associated with the macroalgae species of *Turbinaria decurrens* and *Halimeda* sps. Information on macroalgal species associated with *G. belizeanus* isolated from the Gulf of Aqaba is unavailable [42]. In the case of *G. yasumotoi*, the macroalgal substrates were identified as *Padina tetrastomatica* and *Sargassum oligocystum* in winter (November 2012) collections whereas in summers (March 2013) they occurred along with *Chaetomorpha* sp. The group also reported the co-occurrence of macroalgal species of *Amphidinium carterae*, *Coolia monotis*, *Ostreopsis* sps., *Prorocentrum formosum*, *P. tsawwassenense*, *Peridinium quinquecorne*, *Adenoides eludens,* and *Cabra matta* [42]. Moderate Resolution Imaging Spectroradiometer (MODIS) and Medium Resolution Imaging Spectrometer (MERIS) data and its various operational algorithms such as OC3M-547, GSM, GIOP, and other bio-optical methods developed by Uddin and coworkers [41,45,54], revealed algal blooms; during field verification, it was observed that *G. toxicus* was also part of the algal bloom on 21 December 2009. Samples collected from location 29.38694◦ N, 47.78389◦ E had over 1000 cells/L. Other algae identified in the field samples were *G. toxicus*, *Karenia selliformii*, *K. brevis*, *P. lima*, *Dinophysis rotundata*, *Ceratium tripos*, and *Myrionecta rubrum*.

Recent deliberations by the experts in an international meeting in Rome on ciguatera poisoning recommended the next generation sequencing (NGS) to mass-amplify specific gene sequences from sediment samples to characterize all species or specific taxa present in the sample [8]. This approach allows better identification and resolution of microbial community composition than the conventional morphological and molecular methodologies. The information on ciguatera from the Indian Ocean region is still in its initial phase. This approach would provide comprehensive information on ciguatera diversity and its interaction with associated communities as reported elsewhere [87–91].

#### *3.4. Ciguatera Causative Organism's Morphology and Phylogeny*

A variety of techniques are known to be used for ciguatera identification at a particular site. These include light microscopy (LM), scanning electron microscopy (SEM), DNA sequencing, restriction fragment length polymorphism (RFLP), quantitative polymerase chain reaction (qPCR), and metabarcoding [8,87]. While performing the morphological identification, the taxonomic key reported earlier [25] was followed by the research groups to confirm the species of *Gambierdiscus* found in the region. Earlier in the 1990s, SEM was used for the identification of *G. toxicus* from Reunion Islands [64]. SEM was also used recently to identify the species of *G. toxicus*; *G. belizeanus*; *G. polynesiensis*; *G. australes;* and *F. yasumotoi* [69]. The morphological identification of Gulf strains of *G. yasumotoi and G. belizeanus* was based on both LM and SEM [42–44]. A description of morphological parameters of strains present in the Indian Ocean region is given in Table 2.


**Table 2.** Morphological features of *Gambierdiscus* species found in the Indian Ocean region.

Owing to the plasticity of morphological characteristics of some species of *Gambierdiscus,* the identification can be ambiguous if only morphology is used. Currently, molecular techniques are gaining popularity for the confirmation of species identified through microscopic methods. The only report on molecular identification of *Gambierdiscus* in the Arabian Gulf was by Catania [43] and her team [44]. The group amplified an 850 bp hypervariable (D8-D10) region of the larger subunit (LSU) of rRNA using primers FD8 and RB [11,25] through Sanger sequencing. A query of assembled nucleotides on the National Centre for Biotechnology and Informatics (NCBI) through the basic local alignment search tool (BLAST) resulted in a 100% match with *G. belizeanus*. Their (KY782637–KY782645) strains depicted a close resemblance with the Caribbean strain of *G. belizeanus*. The group also reported a 116 bp deletion at 493–609 position in one of the isolates, probably imparting the distinctiveness from other species of *Gambierdiscus* [95,96].

Phylogenetic analysis (methodology described in the Supplementary file) based on multiple alignments (Clustal2.1) of all the available DNA sequences (n = 345) of the D8–D10 region of the large sub unit (LSU) of ribosomal RNA (rRNA) [9,11,14,15,25,26,40,43,44,89,96–114] distributes the *Gambierdiscus* genus into five major clades (clade I–V) as seen in Figure 8. Clade II and III exclusively contain the species *G. australes* (Atlantic and Pacific strains) and *G. excentricus* (Atlantic strains), respectively. Clade I encompass all of the *G. carribaeus* (Atlantic and Pacific

strains) and newly discovered *G. carpenteri* (Atlantic and Pacific strains) as well as *G. jejuensis* (Pacific strains). Clade IV and V are more diverse. Of this, *G. belizeanus* (Red Sea, Saudi Arabia; Charco Azul, El Hierro, Spain; St. Barthelemy Island, Caribbean Sea) appears as a distinct branch in Clade V with the closest genetic relationship with *G. honu* (Kermadec Islands, Australia). The yet unclassified *Gambierdiscus* ribotype 1 follows the same lineage as of *G. belizeanus*. Other species in close proximity were *G. balechii* (Celebes Sea, Pacific Ocean; Phuket Islands, Indonesia, North Pacific), *G. lapillus* (Great Barrier Reef, Australia, Pacific Ocean; Cook Islands, Rarotonga, North Atlantic), *G. cheloniae* (Rarotonga, Cook Islands, North Atlantic), *G. scabrous* (Japan, South China Sea, North Pacific), and *G. pacificus* (Balearic Islands, Spain, North Atlantic; Cook Islands, Rarotonga, North Atlantic; Marshall Islands, Micronesia, South Pacific). The *G. toxicus* (Indian Ocean) strains are also in the same clade.

**Figure 8.** Phylogenetic associations of globally distributed *Gambierdiscus* and *Fukuyoa* species. A split decomposition algorithm was applied on multiple aligned D8-D10 sequences of LSU of available *Gambierdiscus* species (downloaded from NCBI) and aligned through ClustalW2.

Molecular identification of *F. yasumotoi* on the Kuwaiti shore waters and Gulf of Aqaba have not yet been performed. Similarly, *G. australes*, *G. toxicus,* and *G. polynesiensis* from Manora Channel, Indian Ocean region are pending molecular identification. *F. yasumotoi* isolated from the Pacific [11] and Atlantic region fall in the same branch as that of *F. reutzleri* (North Atlantic and North Pacific Ocean) and *F. paulensis* (North Atlantic and South Pacific Oceans). All the *F. yasumotoi* strains discovered so far showed a similarity of ~80% with the unique ribotype A213 [11,97]. Other species in the same clade are that of *G. carolinianus* (Bermuda, North Atlantic; Bahamas, North Atlantic; Cancun, Mexico, North Atlantic; Caribbean Sea, North Atlantic), *G. silvae* (Canary Island, North Atlantic), *G. polynesiensis* (Cook Islands, Rarotonga, North Atlantic; Pacific Ocean), and unclassified *Gambierdiscus* ribotypes (Curlew Cay, Belize, North Atlantic).

More diversity in clade IV and V including the species discovered from Indian Ocean may be due to a smaller number of samples owing to the recent discovery of these species. The presence of unclassified *Fukuyoa* sps. and *Gambierdiscus* ribotypes in these clades is suggestive of unexplored novel species. In 2016, at the 32nd session of the Codex Committee on Fisheries and Fishery Products, the Pacific Nations raised CP as an issue that is increasingly affecting the tropical and subtropical regions of the Pacific Ocean, Indian Ocean and the Caribbean Sea between the latitudes of 35◦ N and 35◦ S [8]. The strains found in the Indian Ocean and their adjoining seas lie near the strains commonly found in these high-risk regions. Probably, the interconnection between the water bodies and the ballast water is the major source of ciguatera infiltration in the Indian Ocean waters. Given this, extensive monitoring and risk management program should be developed for the unexplored regions and specifically in the Middle East region.

#### *3.5. Vectors of Ciguatera in Indian Ocean Region*

The CP is recognized as a tropical disease but the existence of ciguatoxic fishes is reported globally due to international seafood trade and shipment [115]. According to a recent FAO/WHO report, globally 425 species of fish, especially those inhabiting the coral reefs, have been associated with CP [8]. The most significant toxic fish have been made by barracuda (Sphyraenidae), amberjack (*Seriola*), grouper (Serranidae), snapper (Lutjanidae), po'ou (*Labridae* spp.), jack (*Carangidae* spp.), trevally (*Caranx* spp.), wrasse (*Labridae* spp.), surgeonfish (*Acanthuridae* spp.), moray eel (*Muraenidae* spp.), roi (*Cephalopholis* spp.), parrotfish (*Scaridae* spp.), etc.

The CP cases have mainly been reported in the southwestern region encompassing Comoros, Mayotte, La Reunion, Mauritius, Rodrigues, and Seychelles. The reports from the Northern part of the Indian ocean have been scarce and involve cases in Mangalore and Trivandrum on the southern Indian coast and an outbreak resulting in one fatality from Pakistan, but the fish species involved and their origin could not be confirmed [74,75,116,117]. Few cases of CFP were also reported in Thailand after eating ocean fish particularly sea bass and red snapper [118,119]. One CP record from Egypt exists [120]. There has been no event recorded in the Northern part of the Arabian sea, in countries bordering the Gulf region or the Red Sea. The very first incidence of ciguatera poisoning in the Indian Ocean region was reported in 1993 from Manakara and Madagascar where about 500 persons were infected after consuming ciguatoxic shark. The fatality rate of this outbreak was 20% [121]. In Trivandrum, India, an autochthonous outbreak was reported in 2015 and 2016, due to which nearly 200 workers of a fish factory contracted CFP after eating heads of Red Snappers [75]. In 2013, also severe food poisoning events were witnessed after consuming a bull shark (*Carcharhinus leucas*), resulting in the deaths of 11 people in Madagascar in 2013 [78]. Fish originating from the Indian ocean region were implicated in ciguatera poisoning in other part of the world. The rapid alert system for food and feed (RASFF) created in 1979 by the European Union (EU) successively warned in 2012, 2015, and 2016 about ciguatoxic fishes originating from India and Sri Lanka (RASFF No. 2012-1602, 2015-0088, 2016-0932, 2016-1152, 2016-1155, 2017-1112). In 2016, scientists and regional public health authorities warned the population in Southwest India about the CFP risk caused by the consumption of Red Snappers [122], yet snappers from India were imported by France and distributed to Germany (RASFF No. 2016-0932). The French Poison Control center network reported from 2012 to 2019, 17 events with fish caught from the Indian Ocean. In a recent report from Germany, the rarely occurring CP outbreaks were reported between 2012–2017, the main reason identified as imports of snappers (Lutjanidae) from the Indian Ocean region mainly from countries like India, Indonesia, and Vietnam. The author emphasized that fishes from the Indian Ocean can cause ciguatera, which has been poorly documented [81].

Although incidences of CP have not been reported from the Arabian Gulf yet, however, the fishes such as grouper, snapper, emperors, barracuda, jacks, trevallies, kingfish, and tuna form the part of common catch by local fishermen in Arabian Gulf [123]. The database search in Fish base returned a minimum of 13 (Iraq) to a maximum of 60 (Oman) fishes

associated with CP elsewhere in the world [124] (Figure 9). An increase in ciguatera causative organism could have a major impact on food safety and food security given that those species are highly consumed in the region. The suspicion of a high number of unreported cases is warranted given that ciguatera is difficult to diagnose and not subject to mandatory reporting [125].

**Figure 9.** Number of Fishes associated with CP in Persian Gulf Countries as per the records of Fish Base (accessed on 1 June 2021).

> The Persian Gulf water is the only source of freshwater through seawater desalination to most Gulf countries and an important source of seafood for the region. Previous reports on incidences of harmful algal blooms on drinking water and food safety and of massive fish kills in the region further creates awareness and interest to investigate the occurrence of ciguatera causative organisms and associated toxins, as well as their health implications in the Gulf countries [58].

#### *3.6. Ciguatera Monitoring in Indian Ocean Region*

Risk assessment and monitoring of Ciguatera in the Indian Ocean region is largely lacking with only a couple of previous attempts from la Reunion [47] and Mayotte Island [61]. In 2014, an algal bloom monitoring system was developed for Kuwait's coastal waters. The OC3M (Aqua-MODIS) and OC4E (ENVISAT-MERIS) algorithms most accurately measured chlorophyll-a concentrations in Kuwait bay. Due to the poor temporal resolution and the decommissioning of ENVISAT-MERIS, Aqua-MODIS data was used for continuous observation. Additionally, algorithms such as Generalized Inherent Optical Properties (GIOP), Garver, Siegel, Maritorena Model, OC2, MODIS fluorescence line height, and the MERIS-based NIR-Red algorithms were attempted, which have a lower accuracy when compared to the OC3M algorithm. The OC3M detected the most reported in-situ algal bloom events (19/50) and most accurately measured chlorophyll-a concentration (RMS: 2.42, RMSE: 4.11, Mean Bias: 54.2%). The Aqua-MODIS OC3M was selected as the preferred algorithm to monitor chlorophyll-a concentration and to detect algal blooms in Kuwait bay and surrounding waters [41,45]. Data variables such as sea surface temperature, OC3M, distance to aquaculture, Garver-Siegel-Maritorena (GSM), generalized

inherent optical properties (GIOP), euphotic depth, Secchi disk depth, distance to shore, precipitation, photosynthetically active radiation (PAR), distance to the river, bathymetry, colored dissolved organic matter, wind direction, speed, and precipitation, etc. were also estimated through a multivariate regression model, a hybrid multivariate regression model, an artificial neural network model, and a hybrid artificial neural network model by Uddin and his group [41,45].

A synergistic model was created that combines the GIS-imaging, the different estimates of environmental parameters, and in situ monitoring with traditional toxin analytical methods [41,54]. Sampling epiphytic substrates and analyzing samples using traditional optical microscopy will provide very useful and immediate information and developing ciguatera early warning systems in the region. The molecular methods of qPCR and metabarcoding can be useful but only as a complement not as the basic methodology to estimate *Gambierdiscus* abundance.

#### **4. Conclusions**

Ciguatera research in the Indian Ocean region is limited and fragmented. Although, there are reported cases of CP in the region. The most significant toxic fish in the region were barracuda (Sphyraenidae), amberjack (*Seriola*), grouper (Serranidae), snapper (Lutjanidae), po'ou (*Labridae* spp.), jack (*Carangidae* spp.), trevally (*Caranx* spp.), wrasse (*Labridae* spp.), surgeonfish (*Acanthuridae* spp.), moray eel (*Muraenidae* spp.), roi (*Cephalopholis* spp.), parrotfish (*Scaridae* spp.), seabass, shark, and red snapper to name a few. Since the region exports fisheries, some of the fish originating from India and Sri Lanka in 2012, 2015, and 2016 were implicated in ciguatera poisoning in Europe.

A comprehensive survey of algal substrates in the region complemented with high throughput metabarcoding would provide insights into novel and undiscovered contributors of ciguatera. Although less in numbers, the presence of ciguatera causative organisms in the region cannot be ignored and their interaction with substrate and other microbial species is worthy of further investigation. The development of an early warning system for HABs is very much the need of the hour.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/ 10.3390/toxins13080525/s1, Metadata and percent identity matrix of 345 Gambierdiscus sequences downloaded from NCBI and aligned using Clustal W2.1.

**Author Contributions:** Conceptualization, S.U., N.H.; methodology, N.H., S.U.; software, N.H., M.F.; resources, S.U., N.H., M.-Y.D.B., M.F.; data curation, S.U., M.F.; writing—original draft preparation, N.H., S.U.; writing—review and editing, N.H., S.U., M.-Y.D.B.; visualization, M.F., M.-Y.D.B. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Institutional Review Board Statement:** Not applicable, since this study does not involve humans or animals.

**Informed Consent Statement:** Not applicable as no human subject was involved.

**Data Availability Statement:** Data provided as Supplementary file and within manuscript itself.

**Acknowledgments:** Authors will like to thank the Kuwait Institute for Scientific Research for supporting this study.

**Conflicts of Interest:** The authors declare no conflict of interest. The Kuwait Institute for Scientific Research has no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**


## *Review* **Digital Technologies and Open Data Sources in Marine Biotoxins' Risk Analysis: The Case of Ciguatera Fish Poisoning**

**Panagiota Katikou**

Ministry of Rural Development and Food, Directorate General of Rural Development, Directorate of Research, Innovation and Education, Hapsa & Karatasou 1, 54626 Thessaloniki, Greece; pkatikou@otenet.gr

**Abstract:** Currently, digital technologies influence information dissemination in all business sectors, with great emphasis put on exploitation strategies. Public administrations often use information systems and establish open data repositories, primarily supporting their operation but also serving as data providers, facilitating decision-making. As such, risk analysis in the public health sector, including food safety authorities, often relies on digital technologies and open data sources. Global food safety challenges include marine biotoxins (MBs), being contaminants whose mitigation largely depends on risk analysis. Ciguatera Fish Poisoning (CFP), in particular, is a MB-related seafood intoxication attributed to the consumption of fish species that are prone to accumulate ciguatoxins. Historically, CFP occurred endemically in tropical/subtropical areas, but has gradually emerged in temperate regions, including European waters, necessitating official policy adoption to manage the potential risks. Researchers and policy-makers highlight scientific data inadequacy, underreporting of outbreaks and information source fragmentation as major obstacles in developing CFP mitigation strategies. Although digital technologies and open data sources provide exploitable scientific information for MB risk analysis, their utilization in counteracting CFP-related hazards has not been addressed to date. This work thus attempts to answer the question, "What is the current extent of digital technologies' and open data sources' utilization within risk analysis tasks in the MBs field, particularly on CFP?", by conducting a systematic literature review of the available scientific and grey literature. Results indicate that the use of digital technologies and open data sources in CFP is not negligible. However, certain gaps are identified regarding discrepancies in terminology, source fragmentation and a redundancy and downplay of social media utilization, in turn constituting a future research agenda for this under-researched topic.

**Keywords:** Ciguatera Fish Poisoning; digital technologies; open data; risk analysis; marine biotoxins

**Key Contribution:** The manuscript summarizes the utilization extent of digital technologies and open data sources with regard to risk analysis tasks related to Ciguatera Fish Poisoning and discusses future research required to approach this topic more in depth.

#### **1. Introduction**

The rapid acceleration of digital technologies, evidenced more intensely during the past decade, globally permeates every private and public organization, transforming their daily working practices, at the same time reshaping social interactions and citizens' expectations [1,2]. Digital tools, among which the Internet, social media, mobile computing, big data, data analytics, and numerous others, open up a fascinating world of innovation opportunities with a significant impact on multiple aspects of contemporary societies [1–3]. This overwhelming penetration of information and communication technologies (ICTs) in everyday life is altering the information sharing preconditions and can technically support more collaborative cultures of information production and dissemination, thus shifting the focus from technology itself to strategies for its exploitation [4].

Substantial implementation of digital technologies in government/public sector operations, commonly intersecting with the e-Government concept, entails the use of public

**Citation:** Katikou, P. Digital Technologies and Open Data Sources in Marine Biotoxins' Risk Analysis: The Case of Ciguatera Fish Poisoning. *Toxins* **2021**, *13*, 692. https://doi.org/ 10.3390/toxins13100692

Received: 20 July 2021 Accepted: 28 September 2021 Published: 30 September 2021

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

**Copyright:** © 2021 by the author. 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/).

information systems and the creation of open data repositories, to serve as tools supporting the fundamental principles of transparency, participation and collaboration [4,5]. Efficient incorporation and interoperability of these tools can improve decision-making, by providing policy-formulators with ample data to address complex problems and to design effective public policies in various governmental disciplines [5,6]. The public health sector, and particularly food safety authorities, should be no exception. Digital technologies and open/big data can be of utmost importance in risk analysis processes, the latter considered necessary to proactively refine and optimize food safety and legislation, rather than maintaining a reactive management approach [7,8].

Emergence of new pathogenic microorganisms and the unintentional presence of chemical contaminants constitute major biological and chemical hazards, respectively, significantly challenging global food safety. Natural toxins, and marine biotoxins (MBs) in particular, comprise a distinct type of food hazard, in the sense that they are chemical toxic substances but are of biological origin [8]. MBs are synthesized by specific marine microorganisms, mainly microalgae (usually termed as phytoplankton) and a few bacterial species. Under certain favorable environmental conditions, toxic or harmful algae may proliferate and aggregate to form dense cell assemblages, commonly known as 'harmful algal blooms' (HABs), accompanied by MBs production able to contaminate seafood, resulting in a serious health threat to consumers. MBs could cause severe human intoxications, such as amnesic, diarrheic, azaspiracid, neurotoxic and paralytic shellfish poisonings and Ciguatera Fish Poisoning (CFP) [9]. MBs are in general heat-stable compounds, resistant to common food processing technologies, whereas no antidote exists to reverse their effects in humans [10]. Illness prevention is thus essential to manage MB-related public health risks, with risk analysis being an irreplaceable tool in the arsenal of public authorities pursuing mitigation of HABs' negative impacts [9,10].

Worldwide, CFP is the most prevalent biotoxin-related seafood poisoning, resulting from the consumption of seafood contaminated by its associated toxins, known as ciguatoxins (CTXs) [11]. Despite the significant under-reporting of cases due to a lack of diagnostic methods, CFP is estimated to annually affect approximately 50,000–500,000 people [12]. Historically, this syndrome is mainly encountered in tropical and subtropical areas. In the recent past, however, a geographical expansion of CFP in more temperate areas has been evidenced. Factors such as climate change, but also some anthropogenic activities, are incriminated for altering the geographical distribution of the causative organisms, which are dinoflagellates of the genera *Gambierdiscus* and *Fukuyoa*, as well as the migration patterns of ciguateric fish [13,14]. Additional factors influencing the occurrence of CFP in non-endemic areas are related to international trade and consumption of imported ciguateric fish species in non-tropical areas and/or travelers returning from CFP endemic areas [11,12]. Current reports of CFP in temperate waters of the Canary Islands (Spain) and the Madeira archipelago (Portugal), accompanied by the documented presence of *Gambierdiscus* and *Fukuyoa* spp. in the Mediterranean Sea, constitute CFP being emerging hazard in European waters, thus necessitating the adoption of official policies to manage the potential risks [12,15]. Nonetheless, the European Food Safety Authority (EFSA) highlighted scientific data inadequacy among the reasons hindering the development of appropriate human health protection strategies against CFP [16], whereas the problems of CFP cases' under-reporting and information sources' fragmentation are emphasized by both researchers and policy-makers [12,16]. Taking into account these shortcomings, in combination with the significant health, socioeconomic and socio-cultural impacts of CFP, as well as its increasing emergence in non-endemic areas, improvements in data collection and availability are evidently required at a global level, to allow for more efficient risk monitoring and mitigation [11,12]. In this context, recognizing some of the technological developments able to generate CFP data and assist their dissemination, as well as compiling a roster of openly accessible information sources for CFP, could facilitate the efforts to tackle the weaknesses identified.

Doubtlessly, developments in digital technologies and open data sources amplified the volume of potentially exploitable scientific information for MBs risk analysis purposes. However, bibliographical references on the advancements achieved with the participation of such means in counteracting CFP-related hazards are scattered, whereas, to date, no substantial summary or research study has cumulatively investigated their utilization. The problem addressed in the present work thus relates to examining digital technologies' and open data sources' utilization in CFP research associated with risk analysis tasks. The research question answered in this review, therefore, is "What is the current extent of digital technologies' and open data sources' utilization within risk analysis tasks in the MBs field, particularly on CFP?"

The absence of targeted review articles and scarcity of structured information on the topic necessitated an in-depth literature investigation, within both peer-reviewed publications and grey literature documents to accomplish this study. For the purposes of this review, grey literature is defined according to the Prague definition, "manifold document types produced on all levels of government, academics, business and industry in print and electronic formats that are protected by intellectual property rights, of sufficient quality to be collected and preserved by libraries and institutional repositories, but not controlled by commercial publishers; i.e., where publishing is not the primary activity of the producing body" [17] (p.11), typically including "conference abstracts, presentations, proceedings; regulatory data; unpublished trial data; government publications; reports (such as white papers, working papers, internal documentation); dissertations/theses; patents; and policies & procedures" [18] (para.2).

The structure of the review is as follows: the next section provides brief background knowledge on the concepts of digital technologies, open data sources and risk analysis, viewed from a public health and food safety perspective, to assist in determining the appropriate keywords for the literature investigation. The subsequent two sections describe the research methodology employed and present the bibliographical analysis results. Finally, the findings are discussed, and relevant future research is suggested.

#### **2. Background**

#### *2.1. Digital Technologies*

Digital technologies are broadly defined as "combinations of information, computing, communication, and connectivity technologies" [19]. An initial review of recent research mainly focusing on the public health and food safety contexts, but also at wider level, reveals that concepts such as 'technology', 'digital technologies', 'ICT', 'information technologies', 'digital media' and 'digital tools' are used interchangeably to refer to a broad set of digital devices and applications, such as websites, databases, blogs, online platforms, mobile/wearable devices, mobile phones, social media and the Internet [20–24]. Digital technologies/ICTs are also strongly intertwined with the 'digital transformation' and 'digitalization' concepts. Indeed, 'digital transformation' is defined as the use of digital technologies/ICTs to enable changes and improvements for achieving business and/or organizational goals [25] and 'digitalization' as the sociotechnical process of using digital infrastructures [1]. In this context, the terms 'digital transformation' and 'digitalization' may also constitute relevant keywords for investigating digital technologies utilization, as they are linkable to improvements and changes in work processes of organizations responsible for risk analyses. Digital technologies' proliferation enhances the quality and quantity of daily generated data, creating conditions of information abundance, able to significantly facilitate public authorities' decision- and policy-making processes [19,26].

#### *2.2. Open Data Sources*

Open data refer to "non-privacy-restricted and non-confidential data produced with public money by public and/or private organizations and made available without any usage or distribution restrictions" [6] (p. 258). Open data, frequently termed also as 'public data', can be enriched with data from other sources, resulting in the emergence of large datasets, known as 'big data' [7]. The latter present specific needs for processing, curation, linking, visualization and maintenance, as their sizes overpass common software tools' abilities, whereas value is generated by the combination of different datasets [19,26].

Public policy development frequently relies on 'open data' and 'big data' availability, being indispensable tools for public organizations. Ample open data in diverse formats are stored in repositories on national or international organizations' websites and also can be exploited by other public institutions, thus counteracting unnecessary duplication and associated costs [6]. However, food safety data and information are generally scattered across the food, health and agriculture sectors, with limited interoperability. Consequently, public authorities in charge of food safety-related risk analysis tasks ordinarily resort to multiple open access scientific resources, such as research project websites, online databases, open-access journals, dissertations or other published material, to obtain upto-date technical information. Efficient access to such sources is granted by the growth of digital technologies [27,28]. For the purposes of the present review, the 'open data sources' concept will also extend to 'big data', including those forms of 'open-source' and 'openaccess' scientific data and software freely available in the public domain [7]. Consequently, the search for appropriate information based on the keywords selected, will also encompass results of common Internet search engines, besides the literature databases [29].

#### *2.3. Risk Analysis in Food Safety*

Risk analysis is a powerful science-based tool for reaching sound, consistent solutions to food safety problems. The Codex Alimentarius Commission defines risk analysis in a food safety context as "a process consisting of three components: risk assessment, risk management, and risk communication" [30] (p. 120). More precisely, risk analysis in food safety is a systematic, disciplined decision-making approach, used to estimate human health and safety risks, to identify and implement appropriate measures for risk control, and to communicate with stakeholders about the risks and measures applied [31]. 'Risk assessment' is the science-based component of risk analysis, comprising hazard identification and characterization, exposure assessment and risk characterization. 'Risk management', on the other hand, involves weighing policy alternatives in consultation with relevant stakeholders, according to the risk assessment outcomes and other factors relevant for consumers' health protection, towards selecting appropriate prevention and control options. Lastly, 'risk communication' entails an "interactive exchange of information and opinions throughout the risk analysis process concerning risk, risk-related factors and risk perceptions, among risk assessors, risk managers, consumers, industry, the academic community and other interested parties, including the explanation of risk assessment findings and the basis of risk management decisions" [30,31].

Food safety risk analyses are carried out by national, regional and international authorities, depending on the nature and localization of the specific risk examined [31]. Scientific knowledge on the food issue identified is considered a prerequisite for successful risk analysis; therefore, aggregation of the largest possible appropriate datasets is essential [28]. Strategies to obtain data on food contaminant issues, particularly MBs, require multidisciplinary approaches combining scientific information from fields such as environmental sciences, biology, chemistry, veterinary science, public administration, epidemiology, public health and toxicology. Data collection can present significant difficulties due to frequent gaps identified in information availability; in this context, the exploitation of digital technologies and open/big data sources may catalyze these efforts [10,28,31].

#### **3. Literature Research Method**

The current state of digital technologies and open data utilization in the field of CFP risk analysis was envisaged by a systematic literature review conducted according to previously established principles [29,32]. Three main steps were followed: (i) selecting appropriate keywords and combinations thereof; (ii) choosing source database(s) and running the searches; and (iii) analyzing the results.

The literature review protocol employed is detailed in Table 1. The focus period was set from 2010 to date (mid 2021). The main keywords identified within the background section were divided into five groups, namely, "Digital technologies", "Open data", "Risk analysis", "Biotoxins" and "Ciguatera", according to the concepts comprising the research topic. Each of the keywords from Groups 1–3 was combined with one or more keywords from the remaining two groups to retrieve the articles of interest, utilizing the Boolean operators "AND" and "OR" (on a case basis) to produce more focused results. Searches were performed separately for each combination of keywords and applied to the journals' abstracts, title and keywords, using the Scopus abstract and citation database of peerreviewed literature. All subject areas were selected, due to the multidisciplinary character of this research.


**Table 1.** Literature review protocol (Scopus search).

<sup>1</sup> An asterisk was used as a wildcard symbol to retrieve all possible variations of the relevant search term.

This strategy yielded only one result when the keywords of the "Digital technologies" or "Open data" groups were combined with keywords of the "Biotoxins" or "Ciguatera" groups. A much higher total number of articles was obtained, as expected, when the "Risk analysis" group keywords were looked-up in combination to those of the "Biotoxins" and "Ciguatera" groups. Searches were merged, and after removing the duplicates, 88 articles of multiple types and subject areas remained (Figure 1). The articles' abstracts were carefully read to assess their relevance and to exclude articles containing the selected keywords in another semantic way, shortening down the list to 28 articles. After full-text examination for the presence of appropriate information, more were excluded as "out of topic", with only 11 studies remaining, a rather expectable outcome considering the narrowness of the field and the specialized nature of the research topic. An additional search in the "Pubmed" database, using the same keyword combinations, yielded five further articles. Thereafter, a thorough Google search was conducted, combining in pairs all the above keywords and some additional terms (e.g., database, smartphone, website, satellite imaging, machine learning), to obtain further material from both the scientific and grey literature, such as

press releases, health and fishery authorities' websites, local media, project documents, codes of practice, etc. Finally, reference lists of all selected documents were reviewed to find other articles of interest, whereas their citations in later publications were also evaluated for inclusion in this review [32].

**Figure 1.** Combined search for the "Risk analysis". "Biotoxins" and "Ciguatera" group keywords by (**A**) subject areas and (**B**) types of retrieved documents. "Other" indicates research disciplines not individually mentioned (refer to Table 1 for details).

> Articles considered relevant contained at least one reference to data input for CFP risk analysis or its individual components (assessment, management, communication) obtained by means of digital technologies and/or open data sources, such as websites, databases, software, social media, specific pieces of digital equipment, etc. It is noted that this research only considers digital equipment utilization in terms of mass-market tools, such as computers and portable digital devices (e.g., notebooks, smartphones, tablets); the use of sophisticated analytical equipment, such as liquid chromatographs and mass spectrometers, although largely incorporating digital components (computerized appliances, support PCs and processing software), is beyond the scope of this work. Similarly, statistical analysis software packages, as well as common office-computer software for word processing, spreadsheets creation, etc., are not included in this literature review, as their use is a prerequisite in CFP data generation. In this context, the above strategy resulted in a final list of 38 articles, of which only 19 were openly accessible to the regular public. In the next step, information of relevance was abstracted from the selected documents and contents were analyzed within the identified research concepts' framework, as presented in the following section.

#### **4. Result**

Keywords found in the 38 articles meeting the eligibility criteria are summarized in Table 2. Notably, references connected to digital technologies were fewer than those categorized within the open data sources concept, with 16 and 33 articles, respectively, whereas 11 articles contained keywords of both groups. 'Database' was the keyword most encountered, with 28 articles, while the highest incidence keyword combination was 'website'–'database', with five articles. Further details are provided in Supplementary Materials Table S1.


**Table 2.** Keywords identified in articles meeting the eligibility criteria.

\* Number of studies containing keywords from individual groups cumulatively exceed the total number of selected articles, as some studies contain keywords of both groups.

#### *4.1. Digital Technologies*

Only three results [33–35] were finally obtained using the exact keywords indicated within the digital technologies concept, combined with those related to ciguatera and risk analysis (Table 1); on the other hand, the extended search for specific digital tools retrieved 13 studies containing the terms 'software', 'smartphone' and 'website', as semantic content relevant to the production, processing and/or communication of the data necessary for CFP risk analysis (Table 2). Interestingly, only two articles combining 'social media' and 'ciguatera' within the context of risk analysis were retrieved, despite the existence of several CFP-relevant Facebook and Twitter accounts (Supplementary Materials Table S2) and the popularity of social media [34,35]. The first one referred to social media mechanisms for food/waterborne complaints surveillance and indicated specific social media accounts serving this purpose [34]. The second one only mentioned the appearance of anecdotal reports of CFP cases on social media, such as online fishing for a, where fishers comment on their own experiences providing the opportunity for broader data collection and risk communication, but without pointing to any specific social media accounts [35].

The term 'software' in risk analysis-related CFP studies primarily concerned programs used for molecular/phylogenetic identification of ciguateric fish and CTX-producing microalgae and secondly web applications assisting record-keeping and communication regarding the presence of ciguateric fishes in trade operations [36–38]. Accurate identification of high-risk fish species implicated in CFP and the ability to prevent these from reaching the market, according to regional legislative requirements, are critical in CFP risk assessment, management and communication; therefore, software-based tools can facilitate risk analysis processes [39,40].

Generally, instances of 'website' in the selected articles referred to governmental and organizations' internet pages containing diverse scientific information, including CFP case reports, epidemiological and environmental data, outbreaks occurrence and advice to consumers, as well as other public health data, all being major inputs to CFP risk analysis components [11,39,41–45]. Nevertheless, 'website' was also used by some authors to denote any type of online-available content, such as public databases or even open data portals (Table 3) [41,44,45]. Furthermore, although fishing bans related to geographical origin (known toxic locations), high-risk fish species and fish size restrictions constitute fundamental measures in terms of CFP risk management in endemic areas [11], often communicated to relevant stakeholders through designated websites, social media or applications belonging to public agencies, no relevant articles were retrieved in the literature (scientific or grey) referring to these specific risk communication actions.

Widely marketed digital tools, such as smartphones, have recently emerged as attractive analytical platforms, which in the future may revolutionize food safety control by enabling citizens without any expertise to perform screening tests [46]. A number of

smartphone-based devices or assays have already been developed for various contaminants, including marine toxins [28,46,47] and CTXs, in particular [48]. It should be noted that, currently, smartphones cannot be used on their own to detect food contaminants, without the contribution of some auxiliary part or hand-held device, such as portable electrochemical or optical sensors [28,47–49]. However, they possess independent power sources, computing power, flash-light cameras (i.e., optical systems with constant light sources), web access and wireless data communication, being powerful alternative analytical tools, able to radically change food testing. Although smartphone apps for CTXs testing are not yet commercially available, the future ability of consumers to screen fish for CFP is expected to improve food security and increase public awareness, facilitating also risk assessment and management [47,49].

#### *4.2. Open Data Sources*

Occurrence of keywords belonging to the 'open data sources' group combined to 'ciguatera' was extensively searched, but no studies were found containing 'open data', 'public data' and 'open source', whereas only one publication (a Master's thesis) included the term 'big data' [50]. On the other hand, searching specifically for 'database', after exclusion of instances related to literature/journal databases, resulted in 28 publications containing at least one reference to a data source compliant to the 'open data sources' concept of the present work [11,14,36,37,40,41,43–45,51–66]. Another relevant term encountered in a semantic fitting the concept was 'dataset' [67,68], a term frequently used interchangeably to 'database' [69], while the more general term 'data' was the only one present in other works containing records of CFP incidents derived from public databases [39,70]. A cumulative presentation of the open data sources found in the selected studies is included in Table 3, along with the geographic coverage and an attempt to categorize source types in compliance with the concept description of the present work, using terms as 'open data portal', 'open documents repository', 'public/open source software', etc. This summary is provided in order to explicitly demonstrate the extent, diversity and fragmentation of the available sources, as well as the type of data available for risk analysis purposes, but also to facilitate future CFP research with regard to data retrieval. To our knowledge, all sources included in Table 3 are openly accessible to the regular public, although in some cases a user registration may be required.

The variety of open data source types found in the studied literature (Table 4) indicates that the data derived thereof are sufficiently exploited in the field of CFP research and risk analysis. Evidently though, the terms 'open data' 'public data', 'open source' and 'big data', commonly used in relevant social sciences' research, are practically unknown to authors involved in this field. On the other hand, 'database' was the most frequently used term to describe such information sources, with some articles specifically referring to databases as 'public' [33,55,56], 'web-based' [34], 'online' [59], 'internet' [60], 'electronic' [61] or 'open access' [65], whereas 'online data' was also used in one case [43].



**3.**datausedinCFPliteraturerelatedtorisk(sourcesreferredto'websites'



338






**Reference (2)**

[50] and PW

[68]

[41]

[60] and PW

[41,68] and PW

[14,60] and PW

[14,44,60]

[11]

[68]




*Toxins* **2021**, *13*, 692


343


**Table 4.** Types and total instances of open data sources identified in the 33 selected articles (refer to Table 3 for details).

Geographical coverage of the open data sources found in the selected articles ranged from worldwide to regional, with the majority of non-global coverage sources focusing their data on areas located in the American and Oceania continents, where CFP is long encountered and considered endemic. In contrast, sources targeting for instance European countries, where CFP issue has recently emerged, are scarcer.

The nature of the CFP-related data contained within the identified open data sources varied widely, including data on taxonomy and identification of marine species (fish and microalgae) [11,14,33,36,37,39,43,52,55–57,59–61,66], epidemiology and outbreaks occurrence [11,14,33–35,39–41,43,44,51–55,58,60,61,63–65,70], HAB events [11,14,43,53,60,62–66], climate and environment (temperature, salinity, water quality monitoring, benthic habitats) [33,41,45,50,51,58,62,67,68], public policies and risk mitigation strategies [11,33,43,61] as well as general information on CFP's public health perspective to aid risk communication to the public [11,34,39,41,43,44,51,61,70].

Plurality in open data sources of a similar nature containing data on different regions is also noteworthy, indicating that efforts to collect data, especially those related to CFP surveillance, epidemiology, case reports and outbreaks incidence, are localized and fragmented, even within the same country, such as the data sources of different states within the USA. Conversely, the evident absence of instances of open data sources in certain CFP-susceptible areas of the world, such as some African and Asian countries of the West Indian Ocean, is also notable. Significant redundancies are also encountered, primarily with regard to climate data, and sea surface temperatures in particular, where at least five different sources are available at a worldwide level. Similarly, at least four different open sources exist for fish or algal species taxonomy and identification. As such, policy-makers and researchers undertaking international risk analysis tasks are commonly obliged to resort to multiple information sources and spend considerable time to obtain the required amount of data. On the other hand, discrepancies may also occur between data from different sources, the resolution of which may create an additional burden in order to obtain acceptable data quality for risk analysis purposes.

#### **5. Discussion**

#### *5.1. Research Question Revisited*

This review addressed the research question, "What is the current extent of digital technologies' and open data sources' utilization within risk analysis tasks in the MBs field, particularly on CFP?" Although the commonly expected terminology was almost absent in the relevant bibliography, modifying the search keywords revealed the existence of several CFP risk analysis-related publications, 38 in total, where the data input originated from the use of diverse digital tools and sources. As such, it appears that the current utilization of digital technologies and open data sources in the investigated field is generally not negligible, which reasonably answers the research question.

#### *5.2. Further Remarks on the Findings*

The aforementioned findings demonstrate that exploitation of digital technologies and open data sources in CFP risk analysis and policy-making studies is not negligible, with their utilization being more widespread in scientific works targeting CFP-endemic

areas [40,42,44,45,50–52,60,61,64,66–68,70]. Nevertheless, the pronounced shortage of published works on CFP referring to common social sciences terminology, such as 'digital technologies', 'digital transformation', 'open data' and 'big data', in conjunction with the use of the general terms, e.g., 'website', 'database' and 'dataset', is indicative of an unfamiliarity with these terms regarding the scientific community creating/uploading information and datasets of interest on the Internet, as well as researchers utilizing the data obtained by these sources. Lack of uniformity between the social and natural sciences' terminology is not a new issue; in fact, it forms part of a long-observed general gap between social and natural sciences, thus highlighting the necessity to adopt more transdisciplinary and collaborative approaches across research fields belonging to environmental/marine sciences, toxicology, public health and social sciences [62,71,72].

The fragmented dispersion of data related to CFP surveillance, epidemiology and outbreaks occurrence encountered in the open data sources identified in this literature review, has also been suggested in previous studies. In fact, under-reporting or inconsistent and fragmented recording of CFP cases has been attributed to the absence of formal epidemiological and surveillance methods and a lack of clinical protocols and experience, whereas the need to establish an international register for CFP intoxication cases and also consolidate monitoring of HAB events at a global level is largely emphasized [11,43,53,55,62,73].

Surprisingly only two instances of 'social media' related to CFP [34,35] were found within both peer-reviewed and grey literature publications, suggesting that these digital tools could be underexploited in CFP risk analysis. In fact, food safety agencies already use social media, such as Facebook and Twitter, for risk communication with the general public on food safety issues [26,74], and CFP is no exception. Several CFP-relevant accounts already exist in social media (Table S2), and CFP risk communications, such as notifications of fishing bans or advice to fishers on species and areas at risk, are not uncommon, especially in CFP-endemic regions. On the other hand, online reporting of CFP cases in social media accounts [34], as well as exchange of CFP-related experiences through posts on fishing forums, are also frequent. Evidently, this does not seem to be adequately reflected in the literature, indicating that the impact of social media in the CFP field may constitute a scientific knowledge gap, requiring further research to elucidate their dynamics as dataproviding sources and communication tools in CFP policy-making.

#### **6. Limitations, Conclusions and Future Research**

To our knowledge, to date, no previous works have summarized the utilization of both digital technologies and open data sources in tasks relevant to risk analysis, regarding either MBs or specifically CFP. As such, this review constitutes an initial attempt towards documenting the utilization extent of these tools in CFP risk analysis, according to the currently available literature, but certainly cannot be considered an exhaustive summary of their contribution or an assessment of their effectiveness in this HAB management field. We anticipate this first theoretical approach to trigger further investigation, entailing empirical data, in order to provide concrete evidence on the extent of the interactions between developments in the digital world and their practical applications in the diverse natural sciences fields, including MBs and CFP in particular. In this context, a structured research strategy is required to thoroughly evaluate the impact level of such ICT tools in a qualitative and quantitative way. To achieve this objective, the following approaches are suggested:

(1) Interviewing relevant stakeholders, such as experts, public administrators and researchers, involved in the field of CFP risk management, in order to assess (a) their degree of familiarization with the terminology related to digital technologies and open data sources; and (b) their understanding, own use and perception of specific digital technologies and open data sources. This assessment can be accomplished by means of structured questionnaires, containing both multiple-choice/close-ended (with a rating scale) and open-ended questions, as well as free statements, subsequently followed by content and statistical analysis of the responses obtained. Participation could also be expanded using online forms and/or email-invited questionnaires to more effectively target expert audiences.

(2) Introducing qualitative and quantitative criteria to create a framework for evaluating the impact of the given digital technologies and open data sources and subsequent application of this model to analyze the answers obtained within the context of the available literature. Unequivocally, capitalization of technological progress is the way forward to scientific progress in the modern world. On this basis, accessibility to and exploitation of digital tools and open/big data are synergistically expected to derive innovative applications and services, aiming to facilitate risk analysis and policy-making procedures in the field of food safety, similarly to the progress envisaged in the fisheries sector by the implementation of emerging data technologies, such as blockchain, data mining and artificial intelligence [75]. In the framework of the gaps identified within the present study, research towards consolidation of the currently fragmentary open data sources, such as epidemiological and HAB presence databases, at a worldwide level, can support more robust practices towards mitigation of the CFP problem. On the other hand, embracing the social media potential to strengthen data collection and enhance risk communication channels in the CFP sector is also considered crucial, and definitely requires further scientific research in order to both capture the benefits and tackle the challenges involved. Finally, and most importantly, transdisciplinary collaboration is essential to bridge the evident chasm between humanities and natural sciences, with establishing mutually accepted terminology and definitions for concepts of common interest as a starting point.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/ 10.3390/toxins13100692/s1, Table S1: Main keywords present in the selected articles related to the present review concepts, Table S2: Indicative social media accounts potentially relevant to CFP risk analysis.

**Funding:** This research received no external funding.

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

**Informed Consent Statement:** Not applicable.

**Conflicts of Interest:** The author declares no conflict of interest.

#### **References**


MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Toxins* Editorial Office E-mail: toxins@mdpi.com www.mdpi.com/journal/toxins

MDPI St. Alban-Anlage 66 4052 Basel Switzerland

Tel: +41 61 683 77 34

www.mdpi.com ISBN 978-3-0365-6429-6