Mycotoxins at the Start of the Food Chain in Costa Rica: Analysis of Six Fusarium Toxins and Ochratoxin A between 2013 and 2017 in Animal Feed and Aflatoxin M1 in Dairy Products

Mycotoxins are secondary metabolites, produced by fungi of genera Aspergillus, Penicillium and Fusarium (among others), which produce adverse health effects on humans and animals (carcinogenic, teratogenic and immunosuppressive). In addition, mycotoxins negatively affect the productive parameters of livestock (e.g., weight, food consumption, and food conversion). Epidemiological studies are considered necessary to assist stakeholders with the process of decision-making regarding the control of mycotoxins in processing environments. This study addressed the prevalence in feed ingredients and compound feed of eight different types of toxins, including metabolites produced by Fusarium spp. (Deoxynivalenol/3-acetyldeoxynivalenol, T-2/HT-2 toxins, zearalenone and fumonisins) and two additional toxins (i.e., ochratoxin A (OTA) and aflatoxin M1 (AFM1)) from different fungal species, for over a period of five years. On the subject of Fusarium toxins, higher prevalences were observed for fumonisins (n = 80/113, 70.8%) and DON (n = 212/363, 58.4%), whereas, for OTA, a prevalence of 40.56% was found (n = 146/360). In the case of raw material, mycotoxin contamination exceeding recommended values were observed in cornmeal for HT-2 toxin (n = 3/24, 12.5%), T-2 toxin (n = 3/61, 4.9%), and ZEA (n = 2/45, 4.4%). In contrast, many compound feed samples exceeded recommended values; in dairy cattle feed toxins such as DON (n = 5/147, 3.4%), ZEA (n = 6/150, 4.0%), T-2 toxin (n = 10/171, 5.9%), and HT-2 toxin (n = 13/132, 9.8%) were observed in high amounts. OTA was the most common compound accompanying Fusarium toxins (i.e., 16.67% of co-occurrence with ZEA). This study also provided epidemiological data for AFM1 in liquid milk. The outcomes unveiled a high prevalence of contamination (i.e., 29.6–71.1%) and several samples exceeding the regulatory threshold. Statistical analysis exposed no significant climate effect connected to the prevalence of diverse types of mycotoxins.


Introduction
Mycotoxins are toxic fungal metabolites that can be found in feed ingredients and compound feeds [1,2]. Due to their compositions, they are detrimental to animal and human health [3][4][5][6][7][8]. Currently, more than 400 different types of mycotoxins have been identified [9]. However, Fusarium toxins are among the most commonly monitored as they are acknowledged to present serious health concerns [7,10]. Under certain conditions, some fungi can produce several toxins simultaneously [11][12][13].
In feed production, ca. 60% of the formulation consists of cornmeal, soybean meal, and their derivates [14,15]. In Costa Rica, cereal production represents 38% of the agricultural sector imports [16], where its main suppliers are the United States and Brazil with 84% and 15% contribution, respectively [17]. In this regard, corn imports have increased from 738,539.97 to 781,903.54 metric tons from 2015 to 2017 [18]. On the other hand, soybean imports have risen to 309,897.97 metric tons per year, even though 83% of the soybean meal used as a feedstuff comes from national production [18]. Furthermore, only 38% of the products destined for animal consumption are from national origin, representing a total feed production of 1,238,243 metric tons in 2017. Approximately 45%, 27%, 20%, and 4% of this production is intended to be destined to poultry, higher ruminants, swine, and pets (i.e., cats and dogs), respectively [18]. That is, import and export of animal feed and feed ingredients play an essential part in the co-occurrence of various types of mycotoxins in the finished feed [19,20]. Hence, co-occurrence could be a far more certain and prevalent issue in real mycotoxin feed analysis [11,12,[20][21][22][23].
Mycotoxin metabolites retain toxicity and thus must be surveilled [24,25]. Mycotoxins and their metabolites have several implications for animal and human health. Some are identified/classified as teratogenic, genotoxic, carcinogenic, and immunotoxic. The ingestion of contaminated feed affects animal health and may reduce productivity in animals, generating economic losses [26]. Some mycotoxins ingested and metabolized by productive animals could be accumulated in different organs and tissues reaching the food chain through meat, milk, or eggs [24,27,28]. In Costa Rica, during 2018, consumption of these commodities was estimated in 58.7 kg (i.e., 14.3, 15.4, and 29 kg year −1 for cattle, pork, and chicken, respectively), 215 L, and 218 units per capita, individually [18].
In this regard, epidemiological information tends to be more comprehensive when exploring data from several toxins simultaneously [29]. Accurate mycotoxin data about their presence in feeds are paramount for stakeholders' decision-making process towards the risk management in their manipulation [30]. Numerous reports have explicitly documented the incidence of mycotoxins in feeds, especially in Europe [11,31,32], USA [33], Asia [31], and China [34]. Nowadays, there are insufficient reports oriented to describe the incidence of mycotoxins in feed in Costa Rica. The emphasis has been made towards the investigation of aflatoxins [35,36].
Herein, the prevalent data from feed and feed ingredient samples of eight different toxins, mainly produced by Fusarium spp. (deoxynivalenol/3-acetyldexoynivalenol (DON/3-ADON), T-2/HT-2 toxins, zearalenone (ZEA) and fumonisins (FB 1 and FB 2 )), but also ochratoxin A (OTA), during five years are provided. Finally, in the same period, we analyzed the behavior of AFM 1 in liquid milk.

Mycotoxin Prevalence in Feed Ingredients
In the matter of feed ingredients, cornmeal exceeded guideline values for HT-2 toxin (n = 3/24, 12.5%), T-2 toxin (n = 3/61, 4.9%), and ZEA (n = 2/45, 4.4%) ( Table 2). In a soybean meal, merely HT-2 toxin (n = 1/6, 16.7%) was detected in this situation, and just one sample of wheat had an excessive amount of DON (n = 1/8, 12.5%) ( Table 2). With reference to other raw materials, of less inclusion, such as rice byproducts, palm oil byproducts, of the citrus industry, as well as forages, silages, and hays (treated as a whole group), there are no regulatory guidelines to establish an acceptance parameter. However, it is interesting to notice that, in the groups described above, they share as a common feature a high prevalence of DON (i.e., 66.7%) ( Table 2).    [37] and (2013/165/EU) [34]. c Prevalence is calculated based on the number of samples above limit of detection.
a Toxins detected only once for a specific matrix type were not included. b Plant-derived constituents according to guaranteed labels. Data in parentheses indicate maximum inclusion recommended for each ingredient during feed formulation. † Data compiled from [15,[39][40][41][42]. c Data in parentheses indicate maximum permitted or recommended toxin concentrations according to EU Commission Recommendations (2006/576/EC) [37] and (2013/165/EU) [38]. d Prevalence is calculated considering the number of samples above limit of detection.

Geographical Distribution and Climate Influence for Fusarium Toxins Present in Animal Feed
Geographical and national toxin hotspot distribution was similar for those toxins produced by Fusarium species (Figure 1A-G). A completely different profile was observed when studying OTA and AFM 1 . Interestingly, only 3-ADON and HT-2 toxins prevailed during the rainy season. For other toxins, there were no differences in the levels of contamination between the dry season and the rainy season ( Table 4). As expected, the co-occurrence of two different toxins was the most common situation (i.e., n = 141/279, 50.5%) ( Table 5). Therefore, as the number of simultaneous toxins increased, co-occurrence was less likely to be found (Table 5). In the case of the parent compound-metabolite comparison, the most common combination was the pair T-2/HT-2 toxin with (n = 66/155) 42.6% of prevalence, followed by FB 1 /FB 2 (n = 23/137, 16.8%) and DON/3-ADON (n = 18/177, 10.2%) ( Table 5).

OTA Prevalence in Animal Feeds
Referring to OTA, the total prevalence from 2012 to 2017 was 40.6% (n = 146/360), ranging from 16.3% (n = 8/49) in 2013 to 76.6% (n = 49/64) in 2015. The maximum OTA reported level was 1810 µg kg −1 , in 2016 (Table 1). Only one sample exceeded the maximal advisory level for ochratoxin; this sample corresponded to poultry feed where the recommended concentration is 100 µg kg −1 . The overall OTA prevalence in non-traditional ingredients, poultry, and fish feed was of 56.3%, 44.4%, and 66.7%, respectively (Tables 2 and 3). Furthermore, in May and September, the highest global concentrations of OTA were presented, corresponding to the rainy season releasing an evident difference compared with the findings of the dry season ( Table 4). As the presence of OTA involves other toxin-producing fungi (other than Fusarium), co-occurrence with other metabolites is a possibility. The most prevalent Fusarium toxins present in feed (different from OTA), in decreasing order of incidence, were ZEA, DON + FB 1 , FB 1 , and T-2 toxin with (n = 17/102) 16.7%, (n = 14/102) 13.7%, and (n = 12/102) 11.8% of incidence, respectively ( Table 5). As expected, OTA incidence had a completely different geographical/spatial ( Figure 1H) and thermo/temporal ( Figure 2H) distribution, when compared with the other toxins.

Aflatoxin M 1 in Liquid Milk
Water buffalo milk and butter samples were also analyzed for the presence of Aflatoxin M 1 . Water buffalo (Bubalus bubalis) milk samples (n = 2) were reported below the limit of quantification (i.e., 0.014 µg kg −1 ) and butter (n = 3) ranged from 0.021 to 0.024 µg kg −1 . Even though 2016 was the year with the lowest number of analyzed samples, it was also the year when fewer samples surpassed the 0.05 µg kg −1 threshold (Table 6). An increase in AFM 1 prevalence with 71.1% and 63.2%, respectively (Table 6), was observed during 2014 and 2017. Excluding three samples from 2015, there were no other samples surpassing the US FDA threshold of 0.5 µg kg −1 , thus representing a very small overall percentage for the four years of the study (i.e., n = 3/175, 1.7%). It was studied/monitored that, consistently, higher concentrations of AFM 1 were obtained during March, August, and September (Table 6 and Figure 2I).

Mycotoxin Prevalence between 2013 and 2017 in Animal Feed
Most of the studied toxins (except for 3-ADON, FB 1 , and HT-2) had prevalences higher than 40% during the five years. The average concentrations found in the different toxins in animal feed did not vary between one year and another, except for ZEA and T-2. The drastic increase of ZEA concentrations during 2017 was observed in corn meal and sorghum silo. There is a prior documented avidity of Fusarium spp. to produce ZEA when using moderately alkaline cereals (e.g., maize) as substrates [43]. A general drop in annual temperature may have provoked this upsurge in ZEA contamination. For example, Fusarium graminearum has demonstrated that conditions of pH 9 and incubation temperature of 15.05 • C are required to favor ZEA production [44]. Interestingly, the most toxicologically relevant levels for ZEA were encountered at relatively low temperatures (i.e., near 15 • C). Despite a relatively high prevalence for mycotoxins (i.e., between 46% and 99%, except for FB 1 + FB 2 and DON), the positive samples possessed comparatively low concentrations (Table 1) based on guidance values for mycotoxins in animal feeds within the European Union (see Appendix A Tables A1 and A2) [37,38]. This relatively low toxicological burden could be associated with the control of mycotoxin in animal feed and raw materials that were established in the country since 2007. This control policy covers the majority of the toxins analyzed in this study added to the control of imported raw materials, before its distribution. In coherence to what has been stated, since 2013, proficient manufacturing practices have been evaluated and audited by regulation in animal feed plants. These proficient practices involve the management of raw materials and storage measures, among others, contributing to the reduction of mycotoxin contamination [45].
However, some of the samples were observed with concentrations above the established guidelines with potentially adverse effects on animal health and productivity. It is worth of mentioning the fact that human health could be affected through the consumption of foods of animal origin contaminated with mycotoxins or their metabolites [24,27,28].

Prevalence in Feed Ingredients
Vegetable ingredients may represent from 80% to 100% of the feed (e.g., in ruminants, animal origin ingredients are prohibited) [14,46,47]. For these vegetable-based formulations, corn and soybean meal may represent up to 60% of the input [14,15]. Costa Rican soybean meal and corn, as well as other relevant ingredients, are imported [18]. Quality grain assessment is a degree-based classification. Usually, grade 2 or 3 corn is purchased for feed production [18]. At least 97.9% of the samples contain around 3% of cracked material, and 36.2% of the samples exhibited higher moisture content (i.e., 17%); both factors promote the proliferation of fungi [48]. Toxin-wise, AFB 1, and DON were assayed and are regulated according to FDA criteria. Only 1.9% samples exceeded levels for AFB 1 but none for DON [49]. The data reveal coherence with the obtained results (Table 2). Notwithstanding, a high prevalence for DON was detected and reported by other researchers both for corn and wheat [49]. Conversely, a relatively lower incidence was found in OTA, different from what was conveyed elsewhere [50].

Prevalence in Cattle Feeds
In both dairy and meat cattle, forage, hay, and silage input must not be underplayed, especially in countries where extensive feeding systems based on grazing cattle predominate. Considering Costa Rica a particular case, 85% and 95.9% of the dairy and beef cattle are based on grazing farming, respectively [51]. Relatively favorable toxin profiles were still found in the tested samples. Thereby, surveillance efforts have been focused on compound feed. Generally speaking, ruminants are relatively less sensitive toward the effects of mycotoxins as rumen bacteria play a detoxification role [35,38]. For example, for DON (prevalence of 70.0% and 55.1% in beef cattle feed and dairy cattle feed, respectively), Charmley and collaborators determined that concentrations of 6000 µg kg −1 neither affect feed intake nor are biotransferred to the milk [36,52].

Prevalence in Compound Feed destined for Poultry and Swine
Mycotoxin effects over monogastric animals are varied, depending on the species and physiological and productive stage [53]. For example, in pigs, fumonisin feed contamination is related to pulmonary, hepatic and cardiovascular lesions [54] while DON has been associated with a reduction of productive parameters and feed efficiency [54]. Besides, pigs are especially sensitive to ZEA, as it is directly related to reproductive disorders and low fertility rates [55]. Mycotoxin findings in poultry feed are also worrisome as birds are noticeably susceptible to molecules such as DON. For example, in broilers, trichothecene exposure (e.g., DON), through feed, increases mortality, reduces immune function, and impairs weight gain [56].

Prevalence in Pet Food
Mycotoxins in pet foods have already been reported by other countries, including industrialized ones (e.g., Portugal, USA, England, and Brazil) [57]. Mainly, Fusarium and Penicillium toxins have been described [51]. An elevated prevalence was described for DON and FB 1 (50.0% and 93.3%, respectively) [58]. Mycotoxicosis in pets is associated with chronic disease, liver and kidney damage, and cancer [58]. Finding mycotoxins in thermally treated foods is not uncommon as mycotoxins molecules can withstand relatively elevated temperature; low toxin reduction will occur during extrusion. Fungi colonization of pet extruded food is expected to be low as it possesses relatively low values of moisture and water activity [58,59]. Mycotoxin in pet foods may represent an additional burden to humans due to the pet closeness with their owners.

Prevalence in Fish Feed
Presence of mycotoxins in fish feed is another proof of an industry which has progressively substituted animal protein sources for vegetable ones [60,61]. In this regard, DON, OTA, and ZEA have been said to be responsible for weight loss, exacerbated feed conversion, and increased susceptibility to infection and disease in fish [61,62]. In line with the data reported herein, a recent report revealed that commercial fish feed samples were frequently contaminated with DON (i.e., over 80% of the samples) with mean concentrations of 289 µg kg −1 [49]. Levels as low as 4.5 mg DON kg −1 feed have already confirmed adverse effects in productive parameters and increased mortality in some fish. even in a relatively short period [62].

Geographical Distribution and Climate Influence for Fusarium Toxins Present in Animal Feed
A different spatial distribution profile was observed for AFM 1 and OTA, which are not produced by Fusarium species. Fusarium species have the potential of simultaneously producing the remainder of the toxins assayed [63,64]. OTA is a toxin produced by several fungal species including Aspergillus ochraceus, A. carbonarius, A. niger and Penicillium verrucosum [65]. On the other hand, AFM 1 is not only produced by Aspergillus species but it is also a product of metabolism [66]. Our data not only demonstrate that most sampling weight is centered on the Costa Rican Central Valley plateau, but the largest concentrations also occur therein (geographical zones with a high average relative humidity of 82%). The data also demonstrate that the intricate climate in tropical countries (such as Costa Rica) predicts the behavior of mycotoxin contamination as more challenging.

Aflatoxin M 1 in Liquid Milk
Milk is not only a staple commodity by itself, but it can accompany other potentially contaminated products (e.g., coffee, tea, or chocolate). Additionally, although AFM 1 is the most studied toxin in milk, other toxins have been described as well [67]. Other dairy products are derived from this raw material (e.g., cheese). Although processing is involved, these other dairy products can carry by themselves aflatoxin metabolites as well (see, for example, [68]). During 2017 alone, milk consumption was calculated to be 212 kg per capita [18]. Assuming the worst-case scenario (a sample with the highest concentration of 0.989 µg kg −1 ), a Costa Rican citizen could be exposed up to 210 µg AFM 1 per year. Similarly, a Jersey calf weighing 25-30 kg at birth would be fed with 10% of its live weight with contaminated milk (from 2.5 to 3 kg of milk per day) [69]. Reiteratively, this means a daily exposure of 2.5-3 µg AFM 1 per day. Milk weaning can occur at ten weeks old [70]. Milk consumption level exposure is estimated to be 0.023 ng AFM 1 per kg body weight per day when a maximum level of 0.5 µg kg −1 is used.
Much higher average concentrations of AFM 1 have been documented in other Latin-American countries [71]. Interestingly, AFB 1 (the parent compound of AFM 1 ) has been reported to be present in milk samples [71]). Besides the toxic burden that AFB 1 and AFM 1 have in the liver, recent evidence suggests that kidney toxicity is a certainty [66]. On the other hand, considerably low (i.e., 0.037 µg kg −1 ) AFM 1 levels in milk have been recently reported, although prevalence rates are also relatively high (i.e., 38.8%), [71]. Other Latin-American countries have reported similar percentages [72][73][74][75], and recent prevalence studies have been published in industrialized countries [76][77][78][79]. Epidemiological studies [1] and risk assessment [80][81][82] are paramount to reduce mycotoxin exposure to both humans and animals.
Aflatoxin-contaminated feed must also be monitored to avoid feeding dairy cows with contaminated batches [83]. For instance, the association among most aflatoxin-contaminated feed ingredients and prevalence has been detailed [36,73]. Although the samples reported herein come from a highly industrialized sector, similar prevalence has been reported in fresh milk from small farms [84]. Consistent with our results, the seasonal distribution does not seem to affect AFM 1 prevalence [71], probably because Costa Rica has a tropical climate. In general, Costa Rica has relatively high temperatures (19-30 • C), humidity (60-91%) and abundant rainfall (1400-4500 mm per year) during a great part of the year (i.e., two distinct seasons), in opposition to an Iranian study exhibited a lower prevalence of AFM 1 in bovine milk during spring [85]. Seasonal variations (i.e., during rainy season) were also described for milk from other species (i.e., sheep, goat, and camel) [81]. Other researchers have not documented a clear tendency regarding AFM 1 occurrence during seasons [73]. It has been suggested, however, that climate change can bear an impact on human exposure to aflatoxins and health [85]. Finally, the burden of AFM 1 exposure for a human can be twice as much as breast milk contamination, as has also been well documented [86]. Although some methods for reducing AFM 1 contamination are available [87], pre-and post-harvest strategies are still the most effective strategies [88].

Conclusions
Toxicologically relevant concentrations were found during the five-year survey as some sample concentrations exceeded the regulatory guidelines. Fumonisin and deoxynivalenol feed contamination is worrisome since these toxins have the capacity of being found in significant levels in these matrices, and, in our case, higher levels of toxins are found in the Central Valley of the country. Therefore, surveillance programs should be expanded to the outermost productive regions of the country to suppress sampling bias, if existing any. Thermopluvial conditions do not seem to have a considerable effect on toxin levels, although some metabolites actually seem to behave concurrently. Fusarium metabolites must be stridently monitored as it is clear that contamination in feed and feed ingredients is unfortunately common; this is especially true for fumonisins and T-2. Feed manufacturers, farmers (both in the field and storage facilities) and pet owners alike should be educated as to the proper conditions for food storage to avoid mycotoxin-producing fungal colonization. Toxin metabolite analysis and co-occurrence are paramount for complete surveillance of toxin feeds, and efficiently execute systems for the control and reduction of mycotoxins, as well as their metabolites in feeds. In addition, a strict control of AFM 1 in milk is necessary, because the prevalence of AFM 1 in milk is considerable and several samples exceeded the regulatory thresholds. It must be remembered that milk is the raw material for a wide variety of dairy products (butter, cheese, and yogurt, among others), therefore, the exposure of the population to this mycotoxin is increased.

DON/3-ADON
DONPREP ® (R-biopharm) columns were used for sample extraction. Briefly, 200 mL of purified H 2 O was added to 25 g of test portion. The mixture was dispersed using an Ultra-Turrax ® (T25, IKA Works GmbH & Co, Staufen, Germany) at 8000 rpm. The supernatant was filtered by gravity over an ashless filter paper (Grade 541, Whatman ® , GE Healthcare Life Sciences, Marlborough, MA, USA). Subsequently, an exact 2 mL aliquot from the supernatant was transferred to the IAC column and passed at 1 mL min −1 using an SPE 12 port vacuum manifold (57044, Visiprep™, Supelco Inc., Bellefonte, PA, USA) at 15 mm Hg vacuum. After a washing step using 2× 10 mL water, the columns were left to dry and then four MeOH fractions of 500 µL were passed through the IAC. The total volume recovered was concentrated to dryness under vacuum at 60 • C. The sample was reconstituted with MeOH to 300 µL and transferred to an analytical HPLC conical vial insert (5182-0549, Agilent Technologies, Santa Clara, CA, USA) before injection into the chromatograph.
Gradient mode starting at 80:20 H 2 O, Solvent A/CH 3 OH, Solvent B as per chromatographic conditions. The rest of the program was as follows: at 0.5 min 80% A, at 5.50 min 90% A, at 10 min 90% A, at 11 min 80% A, and at 15 min 80% A. DON and 3-ADON absorption at 220 nm was exploited for detection purposes. Linear calibration curves ranging from 1.25 to 10.00 µg mL −1 were prepared during quantification. The limit of quantification for DON/3-ADON was 10.00 and 40.00 µg kg −1 .

ZEA
Extraction was performed using 100 mL of CH 3 CN/H 2 O 60:40 and an EASI-EXTRACT ® ZEARALENONE IAC (R-biopharm). Isocratic mode using a 40:10:50 CH 3 CN/CH 3 OH/H 2 O mixture at a flow rate of 0.7 mL min −1 was used as per chromatographic conditions. ZEA natural fluorescence (at λ ex = 236, λ em = 464 nm) was exploited for detection purposes. Linear calibration curves ranging from 300.00 to 1200.00 µg L −1 were prepared during quantification. The limit of quantification was 0.072 µg kg −1 .

OTA
Extraction was performed using 100 mL of CH 3 CN/H 2 O 60:40 and an OCRAPREP ® IAC column. OTA elution from column and resuspension after evaporation was achieved using a 98:2 MeOH and acetic acid solution to ensure OTA protonation. Isocratic mode using a 50:50 H 2 O/CH 3 CN mixture using 0.2 mol L −1 trifluoroacetic acid, pH = 2.1 (74564 Millipore Sigma) at a flow rate of 0.7 mL min −1 was used as per chromatographic conditions. OTA natural fluorescence (at λ ex = 247, λ em = 480 nm) was exploited for detection purposes. Linear calibration curves ranging from 2.50 to 40 µg L −1 were prepared during quantification. The limit of quantification was 0.011 µg kg −1 .

AFM 1 in Milk and Butter
AflaStar ® M 1 (Romer Labs Diagnostic GmbH, Tulln an der Donau, Austria) columns were used for sample extraction. An exact 50 mL of raw or processed milk, previously homogenized and filtered by gravity over an ashless filter paper, was transferred to the IAC column. After a washing step using 3 × 10 mL of water, the columns were left to dry and eluted using MeOH and concentrated as described above in 5.4.1. Isocratic mode using a 10:35:55 CH 3 CN/CH 3 OH/H 2 O mixture at a flow rate of 0.6 mL min −1 was used as per chromatographic conditions. AFM 1 natural fluorescence (at λ ex = 365, λ em = 455 nm) was exploited for detection purposes. Linear calibration curves ranging from 0.50 to 2.00 µg L −1 were prepared during quantification. The limit of quantification was 0.014 µg kg −1 .
In the case of the butter samples, the preparation was performed according to the method in [84]. Briefly, 25 mL of aqueous methanol (70 mL/100 mL) was added to 5 g of butter. Afterwards, the solution was extracted by mixing gently for 10 min at room temperature using sonication. The extract was filtered through a paper filter, and 15 mL of distilled water was added to 5 mL of filtered solution. After that, 0.25 mL of Tween 20 were added and dispersed for 2 min, followed by the entire amount of the sample solution (20 mL) passing over the IAC.

Data Analysis
For Tables 1-3, prevalence is expressed as the ratio between the total of assays above the limit of detection and the total of assays performed for each toxin. Descriptive statistics displayed in Table 1 are expressed without considering samples below the limit of detection. Heat maps used in Figure 1 were rendered using ArcGIS Pro v2.2 (Esri TM , Redlands, CA, USA). For each contaminant, Spearman Rank Order tests were applied to assess the association among the toxin concentration and climatic variables (i.e., precipitation, rainy days and temperature). In this particular case, toxin levels below the limit of detection were considered zero for association purposes; this analysis was performed using SigmaPlot 14 (Systat Software Inc., San Jose, CA, USA). Sampling date was linked to mean monthly values and data were retrieved from the closest climatological station to the sampling region. Meteorological data were provided by the Costa Rican National Weather Service (https://www.imn.ac.cr/boletin-meteorologico).