Harmful algal blooms (HABs) negatively affect not only human health, but also that of marine ecosystems. These effects include hypoxia/anoxia events, decreased water clarity, and altered feeding behavior and toxicosis of marine fauna (Zingone and Enevoldsen, 2000). Toxin-producing algal blooms are recurrent off the coast of El Salvador and are predominately paralytic shellfish poisoning events, which are produced by several paralytic shellfish toxin (PST)-producing dinoflagellates, such as Pyrodinium bahamense var. compressum, Gymnodinium catenatum, and Alexandrium spp. These events have caused human intoxication and death, as well as mortality of marine fauna (Espinoza et al., 2013). Globally, exposure to PST toxins has been associated with mortality of marine mammals (Geraci et al., 1989; Van Dolah et al., 2003; Lefebvre et al., 2016), sea birds (Shumway et al., 2003; Shearn-Bochsler et al., 2014), estuarine turtles (Hattenrath-Lehmann et al., 2017), and sea turtles (Maclean, 1975; Licea et al., 2008).

Four of the seven sea turtle species found in the world nest or forage in El Salvador and its waters. All species are considered vulnerable, endangered, or critically endangered by the International Union for the Conservation of Nature (IUCN, 2018) Red List. In 2005, over 200 sea turtles, including green turtles (Chelonia mydas) and olive ridley turtles (Lepidochelys olivacea), were found dead along the coast of El Salvador. The mortality event was attributed to PST based on concentrations found in brain samples which were as high as of 6278.0 μg STX eq kg⁻¹ (Licea et al., 2008). Additionally, in 2010 two hawksbill turtles (Eretmochelys imbricata) were found dying after a Pyrodinium bahamense bloom. One of the turtles was found with 1212.5 μg STX eq kg⁻¹ in brain tissue, and both individuals had diarrhea and exhibited erratic swimming, disorientation, and reduced activity (Licea et al., 2013). However, the limited toxin analytical data prevented confident attribution of clinical signs to PSP (Licea et al., 2013).

Two additional large sea turtle mortality events occurred in El Salvador waters in recent years. In early October 2013, the Ministry of Environment and Natural Resources (MARN) received a report from fishermen of dozens of dead sea turtles floating far from the coast, drifting northwest. In a subsequent report, 40 dead sea turtles were found on beaches of La Paz and La Libertad during September and October, and an undetermined number on beaches in Usulután. On October 28th, 2017, an estimated 400 dead sea turtles were found floating approximately 12 miles off the coast of Bahía de Jiquilisco. In the following days, 160 turtles were observed around 7 miles off the coast of Puerto Parada, of which 60% were L. olivacea and 40% were C. mydas. Between November 3rd and 6th, another 47 sea turtles were found floating off the coast of Usulután and La Paz and 25 L. olivacea individuals were detected 10 miles off El Cordoncillo in La Paz. No other dead marine organisms were observed during either event. The turtles were in good nutritional condition based on muscle mass and body fat and did not have any apparent injuries or other abnormalities to explain the cause of death.

PST-producing HABs were again suspected and samples of tissues, gastrointestinal contents, and seawater were collected for biotoxin analysis and phytoplankton identification. In addition, data collected by satellite (MODIS-Aqua) were reviewed for evidence of possible algal blooms. Herein, we report the results of these analyses as evidence that both mortality events were caused by PST exposure, demonstrating that PST-producing algal blooms are a substantial cause of sea turtle mortality in the region.

Previous sea turtle mass mortality events suspected to be caused by PST-producing dinoflagellate blooms have been reported on the Pacific coast of Mexico and Central America, and in Papua New Guinea; however, very limited information about these events is available (Maclean, 1975; Licea et al., 2008, 2013; Meave del Castillo et al., 2008).

There are significant challenges associated with investigating sea turtle mortality events in many regions of the world, including the remote nature of many coastal areas, limited logistical resources, and inevitable delays between initial observations and response to reports. Although our sample size and some aspects of our data reflect these challenges, we were able to characterize two relatively large sea turtle mortality events in El Salvador and provide evidence that exposure to PST was the most likely cause. This evidence includes postmortem findings consistent with acute toxicosis, detection of relatively high concentrations of PST in gastrointestinal contents and tissues, observation of PST-producing dinoflagellates within gastrointestinal contents, and remote-sensing data suggestive of dinoflagellate blooms concurrent with sea turtle mortality.

In Chelonia mydas, higher concentrations of STX has been found in the gastro-intestinal tract contents rather than tissues (Capper et al., 2013). Toxicity thresholds have not been defined for PST in chelonians or other reptiles. Given their conservation status, exposure studies in sea turtles are not feasible and laboratory studies using non-imperiled chelonians have not yet been conducted. Attribution of clinical effect and mortality currently relies on exclusion of other possible causes and comparison of toxin exposures associated with mortality during known blooms, as well as exposure in turtles that are clinically stable or those that died from apparently unrelated causes. The dead turtles found during both events described herein were consistent with the general features of sea turtle mortality attributed to brevetoxicosis, another biotoxin that acts on neuronal voltage-gated sodium channels (Fauquier et al., 2013; Walker et al., 2018). These features include strandings of turtles in good nutritional condition without injuries, other abnormalities, or known association with other causes of mortality. Similar to a mortality event involving diamondback terrapins (Malaclemys terrapin) associated with a PST-producing Alexandrium bloom (Hattenrath-Lehmann et al., 2017), all turtles were found dead thus we were unable to ascertain whether affected turtles exhibited signs of neurotoxicosis. However, as previously reviewed, abnormal neurological signs including, erratic swimming, disorientation, and reduced activity were observed in stranded turtles found in 2010, one of which had detectable STX in brain tissue (Licea et al., 2013).

The PST concentrations detected in sea turtles that were found during both events were similar to those reported in 2006 (Licea et al., 2008) and 2010 (Licea et al., 2013), the only other published accounts of sea turtle mortality associated with PSP that includes toxin analyses. Our values were much higher than detected in the aforementioned M. terrapin mortality event (Hattenrath-Lehmann et al., 2017). In the case of PST found in one sample from brain tissue in 2013, the value is lower than found in brain tissue from the 2005 to 2006 mortality event (Licea et al., 2008). Two of the authors (Stacy and Flewelling, unpublished data) have screened sea turtles found stranded in Florida for PST as part of health studies and investigation of mortality events. In those results, concentrations typically are below detectable limits (10 ng/g) or well below those documented in this report, including within areas that experience periodic PST-producing blooms. Thus, although some degree of asymptomatic PST exposure almost certainly occurs in sea turtles, we have not found evidence of widespread background exposure similar to that measured during these events as a complication related to interpretation of analytical results. Nonetheless, broader sampling is needed of sea turtles during mortality events and PST-producing blooms and from unrelated circumstances with which to better understand PST exposure in Central America.

The opportunistic nature of sampling during these events did not allow conclusions to be drawn regarding differences in toxin concentrations among turtles and sample types other than variation is attributable to potential differences in dose, timing of exposure, toxin absorption and metabolism, and analytical methods. There are notable differences in the assays performed, which were determined by logistical considerations during each event. For example, RBA measures overall PST toxicity, whereas HPLC measures concentrations of specific congeners (Costa et al., 2009). ELISA is very sensitive and useful for screening for exposure, and analytical confirmation was achieved for the 2017 animals screened by ELISA. However, we did not have the ability to compare results of the same samples across all three methods. Future efforts should endeavor to standardize both sampling and analytical methods to the degree possible.

We observed anomalies in remotely-sensed imagery suggesting possible dinoflagellate blooms during both events. Unfortunately, available water samples were inadequate to confirm that the anomalies observed in satellite imagery were blooms of PST-producing species; corresponding in situ water samples are essential to confirm the presence of a bloom (Hu et al., 2005). The spatiotemporal extent of water sampling was extremely limited and was not considered representative of the anomalies detected by satellite. Nevertheless, the possibility that the observed anomalies were caused by blooms of Pyrodinium bahamanse and the likelihood that such blooms contributed to these mortality events are strengthened by concurrent detection of PST exposure in turtles, detection of Pyrodinium bahamense in the gastrointestinal contents of some turtles, and the observed toxin profiles. PST toxin profiles reported for P. bahamense vary somewhat from location to location, but generally contain fewer toxins than those described for other PST-producing dinoflagellates (Wiese et al., 2010). Profiles consisting of STX, NEO, dcSTX, GTX5, and GTX6 were reported in P. bahamense clones both from Palau (Harada et al., 1982, 1983) and Malaysia (Usup et al., 1994). Pyrodinium bahamense isolates from the Philippines and the southeastern US contained only STX, GTX5, and dcSTX (Landsberg et al., 2006; Gedaria et al., 2007), similar to the PST profile observed in the turtles analyzed in the present study. Although we were unable to confirm and identify blooms of specific species, the remotely-sensed data provided insight into the possible locations and extent of the blooms that may have resulted in toxin exposure and caused sea turtle mortality. Hopefully, this technology will be used to inform real-time sampling during future events.

Another consideration relevant to collection and interpretation of environmental data is the potential for persistent toxin exposure following dissipation of blooms. In marine animal mortality events attributed to brevetoxicosis, animals are not only exposed to high toxin levels during bloom periods, but also to toxins circulating through food webs for weeks or months after a bloom has dissipated (Landsberg et al., 2009). The same potential may exist for PST.

The number of sea turtles found likely reflects a minority of affected animals. Previous studies of beach strandings have estimated that around 5–20% of sea turtles that die at sea are subsequently found on shore (Hart et al., 2006; Mancini et al., 2011; Koch et al., 2013). Therefore, actual mortality from both events could have numbered in the hundreds or even thousands of sea turtles. Moreover, the 2013 event may have been widespread based on the extent of the potential bloom observed in remotely-sensed imagery and reported concurrent sea turtle strandings in Guatemala (Brittain et al., 2014).

Given the status of sea turtles and efforts that are required to manage significant anthropogenic threats, an understanding of various causes of mortality is important for conservation and management. The number of sea turtle deaths attributed to HABs on the Pacific coast of Central America is notable and warrants further effort to understand the factors that contribute to these events.

OA, RQ, and M-YD contributed conception and design of the study. RQ, LF, and BS organized the database. RQ performed the statistical analysis. OA wrote the first draft of the manuscript. RQ, M-YD, BS, RH, LF, CD, and GR wrote sections of the manuscript. All authors contributed to manuscript revision, read and approved the submitted version.