The world's oceans support marine fisheries for commercial, recreational and subsistence uses, and thus are directly linked to human health through fish consumption (i.e. Bergé & Barnathan, Reference Bergé and Barnathan2005; Kite-Powell et al., Reference Kite-Powell, Fleming, Backer, Faustman, Hoagland, Tsuchiya, Younglove, Wilcox and Gast2008; Halpern et al., Reference Halpern, Longo, Hardy, McLeod, Samhouri, Katona, Kleisner, Lester, O'Leary, Ranelletti, Rosenberg, Scarborough, Selig, Best, Brumbaugh, Chapin, Crowder, Daly, Doney Scott, Elfes, Fogarty, Gaines, Jacobsen, Karrer, Leslie Heather, Neeley, Pauly, Polasky, Ris, St. Martin, Stone, Sumaila and Zeller2012; Moore et al., Reference Moore, Depledge, Fleming, Hess, Lees, Leonard, Madsen, Owen, Pirlet, Seys, Vasconcelos and Viarengo2013; Tacon & Metian, Reference Tacon and Metian2013). Fish comprise an important source of animal protein for much of the world's human population, and in the next decade, total production from wild fisheries and aquaculture is expected to exceed production of beef, pork or poultry (FAO/WHO, 2011). In 2010, fish accounted for 16.7% of the world's intake of animal protein, and the world fish food supply grew 3.2% per year from 1961–2012, nearly doubling from an average of 9.9 kg per capita to 19.2 kg per capita (FAO, 2014, pp. 3–4).
The ability of the global population to obtain healthful marine-derived food is dependent on well-managed ecosystems. A broad interdisciplinary approach is needed to understand the connections between the marine environment and human health (Kite-Powell et al., Reference Kite-Powell, Fleming, Backer, Faustman, Hoagland, Tsuchiya, Younglove, Wilcox and Gast2008; Moore et al., Reference Moore, Depledge, Fleming, Hess, Lees, Leonard, Madsen, Owen, Pirlet, Seys, Vasconcelos and Viarengo2013), particularly for evaluating the risks and benefits of consuming seafood. This necessarily requires expertise from marine science as well as public health and biomedical science. This paper is authored by an interdisciplinary group comprising marine and human health scientists who have shared their expertise to synthesize current knowledge on the benefits and risks of consuming marine organisms as routes of human exposure to combinations of fish oils, selenium and the global contaminant mercury, particularly its highly bioavailable and toxic form, methylmercury. Other potential compounds of interest in marine organisms, including organohalogens, natural toxins, arsenicals, trace essential elements and vitamins are beyond the scope of this review; however, introductions to such topics are available elsewhere (Jeandel & Minster, Reference Jeandel and Minster1987; Edmonds & Francesconi, Reference Edmonds and Francesconi1993; Neff, Reference Neff1997; Garthwaite, Reference Garthwaite2000; Lail et al., Reference Lail, Skrabal, Kieber, Bouillon and Wright2007; Guglielmo et al., Reference Guglielmo, Lammel and Maier-Reimer2009; Shaw & Kannan, Reference Shaw and Kannan2009; Yogui & Sericano, Reference Yogui and Sericano2009; Dickey & Plakas, Reference Dickey and Plakas2010; Buck et al., Reference Buck, Franklin, Berger, Conder, Cousins, de Voogt, Jensen, Kannan, Mabury and van Leeuwen2011; Cusick & Sayler, Reference Cusick and Sayler2013; Prego-Faraldo et al., Reference Prego-Faraldo, Valdiglesias, Méndez and Eirín-López2013; Skjånes et al., Reference Skjånes, Rebours and Lindblad2013; Ahrens & Bundschuh, Reference Ahrens and Bundschuh2014; Alonso et al., Reference Alonso, Azevedo, Torres, Dorneles, Eljarrat, Barceló, Lailson-Brito and Malm2014; Sañudo-Wilhelmy et al., Reference Sañudo-Wilhelmy, Gómez-Consarnau, Suffridge and Webb2014). This review provides a limited overview of select dimensions of marine seafood chemical content, and demonstrates the multidisciplinary issues at the interface of Oceans and Human Health (OHH). It does not set out to provide a comprehensive review of seafood content or the overall health implications of seafood consumption. The collaboration of the co-authors of this paper, hailing from diverse disciplinary backgrounds including veterinary medicine, toxicology, immunology, epidemiology, ecology, toxicology and geography, also exemplifies the goals of the OHH initiative which includes the sharing of insights and priorities across research communities (European Marine Board, 2013).
HEALTH BENEFITS OF FISH CONSUMPTION
Fish and shellfish contain protein, long-chain omega-3 fatty acids, vitamins, minerals and trace elements such as calcium and magnesium (Tacon & Metian, Reference Tacon and Metian2013). Seafood has the highest concentration of long-chain omega-3 polyunsaturated fatty acids (PUFAs), including eicosapentaenoic acid (EPA) and docosahexaenoic acid (DHA), of any foods (Tacon & Metian, Reference Tacon and Metian2013). EPA and DHA show beneficial associations with cardiovascular phenotypes including blood pressure (Campbell et al., Reference Campbell, Dickinson, Critchley, Ford and Bradburn2013), vascular endothelial function (Xin et al., Reference Xin, Wei and Li2012), arterial stiffness (Pase et al., Reference Pase, Grima and Sarris2011) and heart rate variability (Xin et al., Reference Xin, Wei and Li2013). Fish or fish oil intake is also associated with decreased weight and waist circumference (Bender et al., Reference Bender, Portmann, Heg, Hofmann, Zwahlen and Egger2014). Possible impacts of EPA and DHA on cholesterol in humans are unclear. Among persons with diabetes, fish oil supplementation may be associated with lower triglycerides and lower levels of very low density lipoprotein (VLDL) cholesterol, but with higher levels of low density lipoprotein (LDL) cholesterol (Hartweg et al., Reference Hartweg, Perera, Montori, Dinneen, Neil and Farmer2008). In dialysis patients, there are also associations of fish oil supplements with lower triglycerides, but also higher high density lipoprotein (HDL) cholesterol, and no association with LDL cholesterol (Zhu et al., Reference Zhu, Dong, Du, Zhang, Chen, Hu and Hu2014). However, the relationship of EPA and DHA to hard cardiovascular endpoints is less clear. A pooled meta-analysis of 68,680 fish oil supplement clinical trial participants, many of whom (more than half of the trials) had pre-existing cardiovascular disease and were being followed for a second event, did not show evidence for lower risk of mortality (from any cause), cardiac death, myocardial infarction or stroke (Rizos et al., Reference Rizos, Ntzani, Bika, Kostapanos and Elisaf2012). In contrast, many observational studies report a decrease in cardiovascular disease and all-cause mortality with higher fish oil intake (Wang et al., Reference Wang, Harris, Chung, Lichtenstein, Balk, Kupelnick, Jordan and Lau2006). The discrepancy between the clinical trial and the observational study results may reflect differences in study populations, or may suggest that another nutrient in fish (or an interacting cofactor in fish) is responsible for some of the cardiovascular benefits attributed to fish oils.
In addition to their possible relevance for cardiometabolic diseases, EPA and DHA fatty acids also may be associated with many other health outcomes. For example, observational studies suggest a lower risk of breast cancer with higher exposure (Zheng et al., Reference Zheng, Hu, Zhao, Yang and Li2013). DHA is essential for ophthalmological and neurological development (Uauy et al., Reference Uauy, Hoffman, Peirano, Birch and Birch2001; Janssen & Kiliaan, Reference Janssen and Kiliaan2014) and fish oil supplements may be associated with cognitive development among infants (Jiao et al., Reference Jiao, Li, Chu, Zeng, Yang and Zhu2014). Among women who previously had delivered a pre-term baby, fish oil supplements appeared to be associated with longer latency and greater weight at birth of the child but did not appear to be associated with differences in risk of another pre-term birth (Saccone & Berghella, Reference Saccone and Berghella2015).
Selenium, present in marine biota including fish and mussels (Outzen et al., Reference Outzen, Tjønneland, Larsen, Andersen, Christensen, Overvad and Olsen2015), has biological effects that are dose-dependent: at low doses, selenium is an essential nutrient used in selenoproteins such as glutathione peroxidase (Barceloux, Reference Barceloux1999), but at higher doses, selenium might be toxic to a range of animals including humans (Barceloux, Reference Barceloux1999; Hoffman, Reference Hoffman2002; Lemly, Reference Lemly2002; Adams et al., Reference Adams, Brix, Edwards, Tear, DeForest and Fairbrother2003; Ackerman & Eagles-Smith, Reference Ackerman and Eagles-Smith2009; Rigby et al., Reference Rigby, Deng, Grieb, Teh and Hung2010; Hladun et al., Reference Hladun, Kaftanoglu, Parker, Tran and Trumble2013; Thomas & Janz, Reference Thomas and Janz2014), although the dose-response of selenium toxicity differs across animal species (Ackerman & Eagles-Smith, Reference Ackerman and Eagles-Smith2009). In humans, the health effects of selenium (total selenium and selenium species) are controversial, with ongoing research into possible elevations or decreases in risk of various health outcomes according to selenium intake (Sabino et al., Reference Sabino, Stranges and Strazzullo2013). A recent Cochrane review (a comprehensive review in medical sciences that aims to summarize published and unpublished data on a topic) of selenium and cancer prevention found heterogeneous studies furnishing no overall evidence that selenium reduces cancer risk (Vinceti et al., Reference Vinceti, Dennert, Crespi, Zwahlen, Brinkman, Zeegers, Horneber, D'Amico and Del Giovane2014).
HAZARDS OF MERCURY
Although seafood provides important nutritional benefits, there may also be hazards from contaminants such as mercury. Neurological impacts of high methylmercury exposure were described in mass poisoning events in Minamata Bay, Japan (Harada, Reference Harada1995) from consumption of seafood contaminated by effluent from a chlor-alkali facility. ‘Minamata disease’ was characterized by deficits in sensation, vision, hearing, coordination (ataxia) and other problems associated with neurological function (Eto et al., Reference Eto, Takizawa, Akagi, Haraguchi, Asano, Takahata and Tokunaga1999; Uchino et al., Reference Uchino, Hirano, Satoh, Arimura, Nakagawa and Wakamiya2005). Children who had high in utero exposures suffered many neurotoxic effects including cerebral palsy, mental retardation, sensorimotor dysfunction and low birth weight (Chapman & Chan, Reference Chapman and Chan2000; Karagas et al., Reference Karagas, Choi, Oken, Horvat, Schoeny, Kamai, Cowell, Grandjean and Korrick2012). At lower doses, the neurological effects of methylmercury are less clear (Axelrad et al., Reference Axelrad, Bellinger, Ryan and Woodruff2007; Karagas et al., Reference Karagas, Choi, Oken, Horvat, Schoeny, Kamai, Cowell, Grandjean and Korrick2012).
Neurodevelopmental toxicity of mercury
Methylmercury neurotoxicity from consumption of seafood has been the focus of birth cohorts in the Faroe Islands, Seychelles and elsewhere (Table 1). In the Faroe Islands, where much of the mercury was acquired from consumption of marine mammals contaminated by organochlorines, there was an inverse association between mercury in cord blood and children's performance on developmental tests (Grandjean et al., Reference Grandjean, Weihe, Burse, Needham, Storr-Hansen, Heinzow, Debes, Murata, Simonsen, Ellefsen, Budtz-Jørgensen, Keiding and White2001, Reference Grandjean, Weihe, Debes, Choi and Budtz-Jørgensen2014). However, in the Seychelles, where much of the mercury was from fish, overall associations between foetal exposure to mercury and neurodevelopmental impairments were generally not observed (Carocci et al., Reference Carocci, Rovito, Sinicropi and Genchi2014). However, at 9 years of age there appeared to be possible differences in fine motor function at higher levels of mercury exposure (Davidson et al., Reference Davidson, Myers, Weiss, Shamleye and Cox2006; van Wijngaarden et al., Reference van Wijngaarden, Beck, Shamlaye, Cernichiari, Davidson, Myers and Clarkson2006; Mergler et al., Reference Mergler, Anderson, Chan, Mahaffey, Murray, Sakamoto and Stern2007), and evidence for interactions between fatty acids and mercury for cognitive processes (Strain et al., Reference Strain, McAfee, van Wijngaarden, Thurston, Mulhern, McSorley, Watson, Love, Smith, Yost, Harrington, Shamlaye, Henderson, Myers and Davidson2015). Emerging research suggests that genetic polymorphisms and epigenetic processes may account for some of the inter-individual variations of health effects given exposures (reviewed in Basu et al., Reference Basu, Goodrich and Head2014). A recent systematic review examined the associations between exposure to methylmercury from seafood consumption and developmental neurotoxicity from 164 studies in 43 countries and found that mercury might be impacting the health of Arctic and riverine populations near gold mining sites, and might also be relevant for public health in highly populated coastal regions (Sheehan et al., Reference Sheehan, Burke, Navas-Acien, Breysse, McGready and Fox2014).
IQR, inter-quartile range (25th and 75th percentiles of distribution).
Immune toxicity of mercury
Data are limited regarding whether mercury from fish affects the immune system, although studies have been conducted in human populations and in toxicological experiments. In cross-sectional studies in Amazonian Brazil, elevated mercury exposures were associated with increased levels of auto-antibodies in gold miners highly exposed to elemental mercury but also possibly exposed to some methylmercury (Silva et al., Reference Silva, Nyland, Gorman, Perisse, Ventura, Santos, de Souza, Burek, Rose and Silbergeld2004; Gardner et al., Reference Gardner, Nyland, Silva, Ventura, deSouza and Silbergeld2010a). A cross-sectional, nationally representative survey of adults in the USA, showed that hair and blood mercury (thought to largely reflect methylmercury exposures) but not urine mercury (thought to largely reflect inorganic exposures) were associated with having anti-nucleolar auto-antibodies (Somers et al., Reference Somers, Ganser, Warren, Basu, Wang, Zick and Park2015). In in vitro toxicological experiments with sufficiently high doses of mercury (3.6 to 36 μM) to induce cell death within 24 h, exposure of human immune cells in vitro to methylmercury prevented B cell proliferation, and these suppressive effects were more severe if mercury exposure occurred prior to immune cell activation (Shenker et al., Reference Shenker, Berthold, Rooney, Vitale, DeBolt and Shapiro1993). In T cells, proliferation was suppressed and apoptosis induced following mercury exposure in vitro, although these effects were examined in mixed culture systems (Shenker et al., Reference Shenker, Berthold, Decker, Mayro, Rooney, Vitale and Shapiro1992; Shenker et al., Reference Shenker, Guo and Shapiro1998). In mixed cultures of peripheral blood mononuclear cells stimulated with lipopolysaccharide, which stimulates macrophages, subcytotoxic concentrations of methylmercury increased production of pro-inflammatory cytokines TNF-α and IL-1β (Gardner et al., Reference Gardner, Nyland, Evans, Wang, Doyle, Crainiceanu and Silbergeld2009, Reference Gardner, Nyland and Silbergeld2010b). Thus, stimulatory effects of methylmercury were observed at doses closer to the typical in vivo human exposure range, generally less than 200 nM (Mahaffey, Reference Mahaffey2004; Mahaffey et al., Reference Mahaffey, Clickner and Jeffries2009), while higher doses were inhibitory. In the more environmentally relevant administered dose studies, effects were primarily observed when cells were stimulated, suggesting that immune activation state at least partially determines the sensitivity to toxic effects on the immune system. If mercury does affect inflammation, then inflammatory mechanisms could impact other organ systems including the cardiovascular system.
Cardiovascular toxicity of mercury
Mercury's potential impacts on the cardiovascular system are a growing area of research (Roman et al., Reference Roman, Walsh, Coull, Dewailly, Guallar, Hattis, Mariën, Schwartz, Stern, Virtanen and Rice2011). Mercury's relationship to fatal heart attacks was recently cited as the potentially most expensive and therefore the most important uncertainty in the cost-benefit analysis for economic benefit of mercury pollution reductions to the USA (Rice et al., Reference Rice, Hammitt and Evans2010). Myocardial infarction and mortality risks from mercury have been evaluated in several recent studies. A cross-sectional survey in a nationally representative sample of South Koreans found a higher odds of previous myocardial infarction with higher levels of blood mercury (Kim et al., Reference Kim, Kim, Yang and Lee2014). A case-control study of 1408 men found that toenail mercury was associated with higher odds of myocardial infarction after accounting for levels of the heart-protective fatty acid DHA (Guallar et al., Reference Guallar, Sanz-Gallardo, van't Veer, Bode, Aro, Gómez-Aracena, Kark, Riemersma, Martín-Moreno and Kok2002). In contrast, a pooled convenience sample drawn from the Health Professionals Follow-up Study and Nurses’ Health Study in the USA (6045 adults) found non-significant, but potentially protective associations between toenail mercury and risk of myocardial infarction, stroke and coronary heart disease (Mozaffarian et al., Reference Mozaffarian, Shi, Morris, Spiegelman, Grandjean, Siscovick, Willett and Rimm2011). This result is acknowledged by the authors to likely reflect the cardio-protective benefits of fish oils, rather than being an accurate portrait of mercury's cardiovascular impact per se. A Swedish cohort also found lower risk of first myocardial infarction with higher erythrocyte mercury, even after controlling for a plasma biomarker of fish oils (Hallgren et al., Reference Hallgren, Hallmans, Jansson, Marklund, Huhtasaari, Schütz, Strömberg, Vessby and Skerfving2001). In contrast, a large cohort study of 1871 elderly men in Finland found strong positive associations of hair mercury levels with acute coronary events, death, and with cause-specific mortality related to congestive heart failure and cardiovascular disease (Virtanen et al., Reference Virtanen, Voutilainen, Rissanen, Mursu, Tuomainen, Korhonen, Valkonen, Seppänen, Laukkanen and Salonen2005). Additional research is needed to clarify whether mercury is causally associated with fatal cardiovascular disease, and to tease apart the reasons for the apparently discrepant findings in the existing literature. It is likely that there are differing distributions of interacting and confounding variables (i.e. other dietary nutrients, or genetics) across these study populations. Data on geographic variation in joint distributions of nutrients and contaminants in seafood would provide important context for interpreting the human health literature.
CONCENTRATIONS OF EPA + DHA
Variability up to 128-fold has been documented in EPA and DHA levels across fish species (Gladyshev et al., Reference Gladyshev, Sushchik and Makhutova2013). EPA and DHA contents in aquatic animals depend on both taxonomic and ecological factors (Makhutova et al., Reference Makhutova, Sushchik, Gladyshev, Ageev, Pryanichnikova and Kalachova2011; Gladyshev et al., Reference Gladyshev, Lepskaya, Sushchik, Makhutova, Kalachova, Malyshevskaya and Markevich2012b; Lau et al., Reference Lau, Vrede, Pickova and Goedkoop2012); other factors such as an anthropogenic pollution (Gladyshev et al., Reference Gladyshev, Anishchenko, Sushchnik, Kalacheva, Gribovskaya and Ageev2012a) may also be important. Research on the possible impacts of fish health status on fish fatty acid content is limited, but suggests the relationships may be complex and organism-specific. In a recent experiment with cultured puffer fish (Fugu rubripes) with or without Trichodina infection, flat fish (Paralichthys olivaceus) with or without streptococcus infection, yellowtail (Seriola quinqueradiata) with or without jaundice, and amberjack (Seriola purpurascens) with or without Photobacterium damselae sp. piscicida, there was not a significant difference by fish disease status in the overall fish fatty acid composition in fish livers; however, liver DHA was significantly higher in the diseased fish than healthy fish for flat fish, yellowtail and amberjack (Tanaka et al., Reference Tanaka, Shigeta, Sugiura, Hatate and Matsushita2014). There is also growing interest in how oxidative stress in fish may affect fish lipids (Tanaka & Nakamura, Reference Tanaka and Nakamura2012; Tanaka et al., Reference Tanaka, Shigeta, Sugiura, Hatate and Matsushita2014).
One objective for our review is to summarize data on EPA and DHA across fish populations. To identify EPA and DHA content of diverse marine fish species, including anadromous fish, we queried Web of Science, Core Collection on 2 October 2014 for ‘fatty acid AND content AND fish AND marine’ (Table 2). Unfortunately, most studies report EPA and DHA as per cent of total fatty acids, and do not provide quantitative information on contents of these PUFA in mass units per fish portion (Gladyshev et al., Reference Gladyshev, Sushchik, Gubanenko, Demirchieva and Kalachova2007, Reference Gladyshev, Lepskaya, Sushchik, Makhutova, Kalachova, Malyshevskaya and Markevich2012b; Huynh & Kitts, Reference Huynh and Kitts2009). In this manuscript, we review data from 10 studies reporting direct measurements of EPA and DHA contents in wild fish biomass obtained using internal standards in chromatography (using capillary columns) over two recent decades. These had slightly different methodologies. For small fish, less than 35 cm (e.g. sardine or capelin), the fish were analysed whole (Huynh & Kitts, Reference Huynh and Kitts2009). Larger fish species (e.g. salmon) were sampled by dissecting muscle tissue (filets without skin), usually under dorsal fin (e.g. Gladyshev et al., Reference Gladyshev, Sushchik, Gubanenko, Demirchieva and Kalachova2006, Reference Gladyshev, Sushchik, Gubanenko, Demirchieva and Kalachova2007, Reference Gladyshev, Lepskaya, Sushchik, Makhutova, Kalachova, Malyshevskaya and Markevich2012b; Huynh & Kitts, Reference Huynh and Kitts2009; Kitson et al., Reference Kitson, Patterson, Izadi and Stark2009; Abd Aziz et al., Reference Abd Aziz, Azlan, Ismail, Alinafiah and Razman2013; Sahari et al., Reference Sahari, Farahani, Soleimanian and Javadi2014). In some studies (Chuang et al., Reference Chuang, Bulbul, Wen, Glew and Ayaz2012) ventral muscles were sampled. In other studies both small and large fish were taken whole, e.g. ground and homogenized (Castro-Gonzalez et al., Reference Castro-Gonzalez, Maafs-Rodriguez, Silencio-Barrita, Galindo-Gomez and Perez-Gil2013). Some authors did not report the sampling in detail (Garcia-Moreno et al., Reference Garcia-Moreno, Perez-Galvez, Morales-Medina, Guadix and Guadix2013).
The resulting data set includes 63 fish species across 11 orders (Table 2). Since PUFA contents in aquatic animals are known to depend on both phylogenetic and ecological factors (Makhutova et al., Reference Makhutova, Sushchik, Gladyshev, Ageev, Pryanichnikova and Kalachova2011; Gladyshev et al., Reference Gladyshev, Lepskaya, Sushchik, Makhutova, Kalachova, Malyshevskaya and Markevich2012b; Lau et al., Reference Lau, Vrede, Pickova and Goedkoop2012), fish species were organized by their EPA and DHA values within taxonomic orders. Putative effects of ecological (habitat) factors were taken into account by dividing the fish species into pelagic, benthopelagic and demersal, as well as by category of water temperature of their habitat, i.e. cold-water, temperate and warm-water (tropical) species. Common size of the fish species was used as a proxy of their trophic level, although this is an imperfect surrogate.
Values of EPA + DHA concentration in the 63 fish species varied from 25.60 mg g−1 (sardine Sardinops sagax, order Clupeiformes) to 0.04 mg g−1 (spotted weakfish Cynoscion nebulosus, order Perciformes) (Table 2). Clupeiformes had the highest median and maximum values of EPA + DHA contents, followed by Salmoniformes, while Perciformes, Scorpaeniformes and Gadiformes and miscellaneous had nearly similar median values (Figure 1). Nevertheless, ranges of values for EPA + DHA content of all orders overlapped in minimum values (Figure 1, Table 2). Thus, all orders, including Clupeiformes, have species with comparatively low content of EPA and DHA.
Interpretation of these results may be complicated by measurement error introduced by differing methods used for fish sampling and analysis, but some broad patterns in the data are interesting. Analysis of published EPA + DHA values found no statistically significant effect of type of habitat (pelagic, benthopelagic and demersal), or temperature of habitat, or their interaction on the PUFA content in fish. To visualize the results of ANOVA, a two-dimensional graph of the PUFA content in the groups of species was created (Figure 2). Since EPA + DHA contents in benthopelagic species overlapped completely with those of pelagic and demersal species, they were not included in the depicted groups. In addition, there were only six cold water species amongst pelagic, benthopelagic and demersal, which were joined in one group. The graph illustrates that EPA and DHA values of all the groups, pelagic temperate water, pelagic warm water, demersal temperate water, demersal warm water and cold water species, overlapped nearly completely.
This analysis of available data did not identify a strong predictor for EPA and DHA contents in fish. Temperature, for example, had limited impact: the contents of EPA + DHA in three pelagic planktivorous Clupeiformes with nearly identical sizes: sardine Sardinops sagax from temperate waters, shad Hilsa macrura from warm waters and herring Clupea harengus from cold waters were all similar (Table 2). Moths et al. (Reference Moths, Dellinger, Holub, Ripley, McGraw and Kinnunen2013) analysed freshwater fish from the Great Lakes as well as 99 other species from freshwater and marine systems documented in seven other studies. As in this study, Moths et al. (Reference Moths, Dellinger, Holub, Ripley, McGraw and Kinnunen2013) found that for marine systems, there was no relationship between latitude and omega-3 fatty acid composition of fish. However, in temperate climates, marine fish had higher omega-3/6 ratios than freshwater fish and for freshwater fish alone, there were higher omega-3 fatty acids in temperate fish as compared with tropical fish. While this study was based on relatively few datasets and many different species, it suggests some interesting patterns. For marine zooplankton, Kattner & Hagen (Reference Kattner, Hagen, Arts, Kainz and Brett2009) did not find significant differences in latitudinal distribution of EPA and DHA levels. Since zooplankton are the main food of these three planktivorous fish species from different latitudes, Kattner & Hagen's (Reference Kattner, Hagen, Arts, Kainz and Brett2009) findings for zooplankton are consistent with those for the planktivorous fish. Thus, more specific characteristics of diverse aquatic ecosystems, such as levels of primary production of PUFA and the efficiency of their transfer through trophic chains (Gladyshev et al., Reference Gladyshev, Sushchik, Anishchenko, Makhutova, Kolmakov, Kalachova, Kolmakova and Dubovskaya2011), are likely to be contributing factors for EPA and DHA content of given fish species. In these large meta-analyses, many environmental and fish specific variables may obscure the potential effects of individual environmental factors such as temperature or trophic level, or pharmacokinetic compartment differences of lipids across fish tissues. More research directed to effects of fish phylogenetics, ecological niche, type of habitat, food quality and other possible determinants is needed to be able to predict EPA and DHA contents, particularly in marine fish.
Studies of fish from field sampling, particularly with heterogeneous methodology, are not conducive to investigating the mechanistic sources of difference between populations living in different environmental settings. In contrast to the analysis of metadata for fish fatty acids above, experimental laboratory studies suggest that fatty acid concentrations in plankton and fish may be influenced in part by the food and temperature environments to which they are exposed. Numerous studies have shown that EPA and PUFAs increase in cells grown at lower temperatures and saturated fatty acids decrease (Thompson et al., Reference Thompson, Guo, Harrison and Whyte1992; Jiang & Gao, Reference Jiang and Gao2004; Fuschino et al., Reference Fuschino, Guschina, Dobson, Yan, Harwood and Arts2011; Teoh et al., Reference Teoh, Phang and Chu2013). In addition, some fish either naturally occurring or cultured have higher concentrations of fatty acids when grown in colder temperatures. Fish need to adjust membrane fluidity for metabolic function in fluctuating temperatures (homeoviscous adaptation) and they do this by changing the concentrations and composition of individual fatty acids and sterols in cell membranes (Sinensky, Reference Sinensky1974; Snyder et al., Reference Snyder, Schregel and Wei2012). Several experimental studies show differences in fatty acid concentrations in fish exposed to different temperatures. Experiments with juvenile Atlantic salmon at two temperatures (14 and 19°C) found that n-3, n-5 and total fatty acids were higher in fish raised in colder water (Arts et al., Reference Arts, Palmer, Skiftesvik, Jokinen and Browman2012). Another study on cultured Atlantic salmon found that the temperature effect was dependent on the type of oil in their food; temperature effects were more pronounced in fish fed copepod oil diets than fish oil diets (Bogevik et al., Reference Bogevik, Henderson, Mundheim, Olsen and Tocher2011). Another study found the digestibility of the lipids in Atlantic salmon to increase with increasing rearing temperatures suggesting that while colder temperatures may favour higher fatty acid concentrations, they may be less digestible than at warmer temperatures (Huguet et al., Reference Huguet, Norambuena, Emery, Hermon and Turchini2015). Laurel et al. (Reference Laurel, Copeman and Parrish2012) found that lower temperatures also favoured increases in unsaturated fatty acids in newly hatched Pacific cod larvae but relative amounts of essential fatty acids did not change with temperature. Similarly, n-3 and n-6 fatty acids decreased with increased temperatures in eggs and larvae of the marine fish, Inimicus japonicas (Wen et al., Reference Wen, Huang, Chen and Feng2013). Thus, there are a range of experimental studies supporting the role of temperature and potentially diet determining fatty acid composition in aquatic plankton and fish. They suggest that colder temperatures result in higher amounts and differing quality of fatty acids. However, the disparity between patterns observed in experimental and field-based studies should be further investigated.
VARIABILITY IN FISH MERCURY CONCENTRATIONS
One of the major challenges in managing human exposure to mercury from fish consumption is that fish mercury concentrations are highly variable. Numerous studies have measured broad differences in mercury content across different finfish and shellfish taxa (Sunderland, Reference Sunderland2007; Karimi et al., Reference Karimi, Fitzgerald and Fisher2012). A recent review estimated that mercury content within a given taxon can also be highly variable, ranging from 0.3–2.4 orders of magnitude, depending on the taxon (Karimi et al., Reference Karimi, Fitzgerald and Fisher2012). This variability poses a challenge to estimating mercury exposure from seafood consumption, and makes it difficult to quantify the risk associated with consuming specific fish taxa.
Numerous studies have shown that body size, age, trophic level and food source of fish are related to concentrations of methylmercury and the per cent of total mercury that is methylmercury (Chen et al., Reference Chen, Dionne, Mayes, Ward, Sturup and Jackson2009; Piraino & Taylor, Reference Piraino and Taylor2009). Across species, body size can be more strongly correlated with mercury concentration than trophic level (Karimi et al., Reference Karimi, Frisk and Fisher2013). In general, larger fish across and within species have higher concentrations of methylmercury because larger fish eat higher trophic level prey and are older and have had a longer time to accumulate mercury (Cossa et al., Reference Cossa, Harmelin-Vivien, Mellon-Duval, Loizeau, Averty, Crochet, Chou and Cadiou2012; Storelli and Barone, Reference Storelli and Barone2013). However, some studies have found that mercury concentration is more strongly correlated with age than length or weight (Braune, Reference Braune1987; Burger & Gochfeld, Reference Burger and Gochfeld2011). For example, the size of Bluefin tuna is not related to mercury concentration (Burger & Gochfeld, Reference Burger and Gochfeld2011) and Atlantic herring in the Arctic show relationships at 3–5 years of age but a decrease at 1–2 years of age due to growth dilution (Braune, Reference Braune1987). While there are clear positive relationships between total mercury and fish size and fish age, there is still variability in total mercury concentrations that is not explained by those two variables as well as the presence of interspecific and intraspecific variability (Tremain & Adams, Reference Tremain and Adams2012). Some of this unexplained variability likely comes from the food source and geographic range of the fish. Fish that have more pelagic than benthic food sources appear to bioaccumulate higher concentrations of mercury (Power et al., Reference Power, Klein, Guiguer and Kwan2002; Chen et al., Reference Chen, Dionne, Mayes, Ward, Sturup and Jackson2009; Karimi et al., Reference Karimi, Frisk and Fisher2013). Not surprisingly, fish that are exposed to higher water and sediment concentrations also have higher tissue concentrations of mercury (Lowery & Garrett, Reference Lowery and Garrett2005; Chen et al., Reference Chen, Dionne, Mayes, Ward, Sturup and Jackson2009; Gehrke et al., Reference Gehrke, Blum and Marvin-DiPasquale2011; Taylor et al., Reference Taylor, Linehan, Murray and Prell2012; Chen et al., Reference Chen, Lai, Chen, Hsu, Hung and Chen2014). However, levels of mercury may vary between similar species in a small geographic area and by tissue within a fish (Bank et al., Reference Bank, Chesney, Shine, Maage and Senn2007). A recent study also suggests increases in methylmercury bioaccumulation in fish experiencing warmer temperatures (Dijkstra et al., Reference Dijkstra, Buckman, Ward, Evans, Dionne and Chen2013). Overall, these studies show that fish size, age, trophic level, food source and geographic region each influence fish mercury content, with no strict rules for which of these factors explains the largest portion of mercury variability. While agencies such as the Food and Drug Administration (FDA) in the USA monitor mercury in marine fish consumed by humans, they do not report fish sizes or geographic location, both of which are extremely important when looking at mercury bioaccumulation.
SELENIUM AND MERCURY CONCENTRATIONS IN FISH
There is a long-running interest in nutrient-toxicant interactions between mercury and selenium (Ganther et al., Reference Ganther, Goudie, Sunde, Kopicky, Wagner, Oh and Hoekstra1972). Although recent evidence suggests possible synergistic interactions between mercury and selenium for fish development (Penglase et al., Reference Penglase, Hamre and Ellingsen2014), the weight of evidence suggests antagonistic interactions in which selenium mediates mercury toxicokinetics (reviewed in Peterson et al., Reference Peterson, Ralston, Whanger, Oldfield and Mosher2009). Selenomethionine increases mercury elimination in zebrafish (Danio rerio) (Yamashita et al., Reference Yamashita, Yamashita, Suzuki, Kani, Mizusawa, Imamura, Takemoto, Hara, Hossain, Yabu and Touhata2013; Amlund et al., Reference Amlund, Lundebye, Boyle and Ellingsen2015), shrimp (Bjerregaard & Christensen, Reference Bjerregaard and Christensen2012) and goldfish (Carassius auratus) (Bjerregaard et al., Reference Bjerregaard, Fjordside, Hansen and Petrova2012); selenite, and seleno-cysteine also increased mercury elimination in goldfish and shrimp. In humans, dietary organic selenium can increase mercury elimination (Li et al., Reference Li, Dong, Chen, Li, Gao, Qu, Wang, Fu, Zhao and Chai2012). Ralston and colleagues report that selenium not only ameliorates the toxic effects of methylmercury by sequestering methylmercury and reducing its bioavailability to organisms, but mercury and selenium may also have physiologically important interactions mediated by other mechanisms (Ralston et al., Reference Ralston, Blackwell and Raymond2007; Ralston & Raymond, Reference Ralston and Raymond2010). Based on rat data, Ralston (Reference Ralston2008) suggests that where the selenium to mercury molar ratio exceeds 1:1, there is adequate selenium to counter mercury toxicity. However, this has not been clearly demonstrated in humans. In recent trout (Salmo trutta) studies in a Norwegian lake, the selenium to mercury molar ratio was a better predictor of trout metallothionein levels than was either selenium or mercury (Sørmo et al., Reference Sørmo, Ciesielski, Øverjordet, Lierhagen, Eggen, Berg and Jenssen2011). However, human studies and clinical trials for selenium demonstrate mixed and inconclusive results for cardiovascular effects of methylmercury and selenium (Mozaffarian, Reference Mozaffarian2009). It has been suggested that mercury cardiovascular toxicity may be modified by selenium intake (Cooper et al., Reference Cooper, Rader and Ralston2007; Khan & Wang, Reference Khan and Wang2009; Mozaffarian, Reference Mozaffarian2009). This might arise through selenium impacts on mercury kinetics (Huang et al., Reference Huang, Strathe, Fadel, Johnson, Lin, Liu and Hung2013) or through impacts on oxidative stress mediators of mercury toxicity (Kaneko & Ralston, Reference Kaneko and Ralston2007; Ralston et al., Reference Ralston, Blackwell and Raymond2007; Farina et al., Reference Farina, Aschner and Rocha2011; Alkazemi et al., Reference Alkazemi, Egeland, Roberts Ii, Chan and Kubow2013; Drescher et al., Reference Drescher, Dewailly, Diorio, Ouellet, Sidi, Abdous, Valera and Ayotte2014), although evidence for the oxidative stress mediation hypotheses is ambiguous (Belanger et al., Reference Belanger, Mirault, Dewailly, Berthiaume and Julien2008). Selenium-mercury interactions may also be relevant for neurodevelopmental outcomes (Choi et al., Reference Choi, Budtz-Jørgensen, Jørgensen, Steuerwald, Debes, Weihe and Grandjean2007).
In recent years due to the interest in selenium to mercury molar ratios, a number of studies have assessed mercury and selenium concentrations and the selenium to mercury molar ratios for a variety of fish species from field samples as well as fish purchased from supermarkets (Burger et al., Reference Burger, Stern and Gochfeld2005, Reference Burger, Gochfeld and Fote2013; Burger & Gochfeld, Reference Burger and Gochfeld2011, Reference Burger and Gochfeld2012; Gochfeld et al., Reference Gochfeld, Burger, Jeitner, Donio and Pittfield2012; Karimi et al., Reference Karimi, Frisk and Fisher2013, Reference Karimi, Fisher and Meliker2014). The relationship between body size and selenium to mercury molar ratios vary with species, tissues and geographic location. Selenium to mercury molar ratios decreased with size of fish for yellowfin tuna and windowpane flounder in Delaware Bay and a wide variety of species in the Aleutians (Burger & Gochfeld, Reference Burger and Gochfeld2011, Reference Burger and Gochfeld2012). Some individuals of most of the 15 species studied in the Aleutians had ratios less than 1.0, where older, larger, higher trophic level fish had the lowest ratios. This was the result of mercury concentrations increasing with fish size but selenium concentrations not increasing with size. While selenium to mercury molar ratios were negatively correlated with fish length for bluefish, the ratios were lower for white muscle tissue, the portion of the fish that humans consume. In a study of 19 species off the coast of New Jersey (USA), (Burger & Gochfeld, Reference Burger and Gochfeld2011) mercury and selenium were positively related for five species and negatively related for two species, and across all species, selenium had no consistent relationship with length. However, for most species tested across all of these studies, the ratios were greater than 1.0, although 20% of the striped bass caught by trawling off the New Jersey coast had molar ratios of less than 1.0 (Gochfeld et al., Reference Gochfeld, Burger, Jeitner, Donio and Pittfield2012).
In general, studies of selenium to mercury molar ratios have found that mercury concentrations were positively related to fish length and trophic level but selenium concentrations were not, and selenium to mercury molar ratios are more strongly related to mercury content than selenium content (Karimi et al., Reference Karimi, Frisk and Fisher2013). This reflects the fact that mercury more strongly accumulates in the body, and biomagnifies through the food chain compared with selenium (Karimi et al., Reference Karimi, Frisk and Fisher2013). These findings are consistent with lower efflux (loss) rates of methylmercury than selenium, because lower efflux rates lead to greater bioaccumulation over time as body size increases (Karimi et al., Reference Karimi, Fisher and Folt2010). However, bivalves (e.g. clams, mussels and oysters) are known to be relatively efficient selenium accumulators (Stewart et al., Reference Stewart, Luoma, Schlekat, Doblin and Hieb2004; Presser & Luoma, Reference Presser and Luoma2010), and have higher selenium concentrations than finfish (Karimi et al., Reference Karimi, Frisk and Fisher2013). It also appears that the mean selenium to mercury molar ratio declines with mean size of fish species and with individual fish size within a species. Both suggest that larger, predatory fish as well as the largest individuals of many species have lower selenium to mercury molar ratios and may not provide selenium protection against mercury toxicity for human seafood consumers (although selenium may be available in their diet from other sources). Moreover, smaller fish of a given species may provide greater protective benefits suggesting that those age classes that reside in more estuarine and coastal environments may present lower human health hazards (Burger et al., Reference Burger, Gochfeld and Fote2013). However, the variability of selenium to mercury molar ratios found within and across species makes it difficult to use this ratio in risk assessment, risk management and risk communication at the present time. Most governmental organizations that develop fish consumption advisories do not have the data on both mercury and selenium levels in individual fish which are necessary to determine the selenium to mercury molar ratio variation within and across species. It is difficult for risk assessors to develop advisories that are protective without an estimate of this variability.
FISH THAT OPTIMIZE POTENTIAL BENEFITS VS RISKS
Recent research is beginning to address the need to quantify the overall nutritional and toxicological value of different types of fish and shellfish based on concentrations of multiple nutrients and contaminants in edible tissues. A recent study found unique, relative concentrations of mercury, omega-3 fatty acids, and selenium, or mercury-nutrient signatures, across seafood taxa (Figure 3, Karimi et al., Reference Karimi, Fisher and Meliker2014). Specifically, salmon and forage fish (herring, anchovies and sardines) are high in EPA and DHA compared with other seafood (Figure 3). In contrast, predatory fish have higher mercury concentrations than lower trophic level fish but nutrient concentrations do not appear to differ as strongly by trophic level. Karimi et al. (Reference Karimi, Fisher and Meliker2014) found that these distinct mercury–nutrient signatures were reflected in the blood of seafood consumers based on their consumption habits. Most notably, consumers with a salmon-dominated diet had a high percentage of omega-3 fatty acids in their blood compared with other seafood consumers. Consumers who tended to eat top-predator fish had higher mercury, but similar nutrient concentrations in blood compared with consumers of lower trophic level seafood. These results suggest that consuming lower trophic level seafood can minimize the risk of mercury exposure without reducing the benefits of nutrient intake, and more broadly, demonstrate the value of examining nutrient and mercury exposure patterns simultaneously. Such research efforts are valuable in summarizing the largest signals among otherwise complex patterns of multiple nutrients and contaminants, but there is a need for a deeper understanding of these multivariate patterns at higher levels of taxonomic resolution. In some cases, the seafood categories used in this study include multiple species that share market names in order to compare mercury–nutrient signatures between edible seafood and seafood consumers. For example, salmon includes Atlantic salmon and multiple species of Pacific salmon, and tuna steak includes bigeye and yellowfin tuna (Karimi et al., Reference Karimi, Fisher and Meliker2014). Future studies that examine the composition of individual fish of the same species would complement these broader analyses by examining nutrient-contaminant patterns at greater taxonomic resolution, and in relation to ecological and environmental factors. In addition, better information on the taxonomic identity of market fish and shellfish would improve estimates of co-exposure to nutrients and contaminants in seafood consumers.
Advice describing both the types and amounts of seafood consumption, while complex, is necessary to better manage risks and benefits of seafood consumption (Oken et al., Reference Oken, Wright, Kleinman, Bellinger, Amarasiriwardena, Hu, Rich-Edwards and Gillman2005; Gerber et al., Reference Gerber, Karimi and Fitzgerald2012). Seafood risk communication also requires risks and benefits to be considered together for appropriate context (Kuntz et al., Reference Kuntz, Ricco, Hill and Anderko2010; Turyk et al., Reference Turyk, Bhavsar, Bowerman, Boysen, Clark, Diamond, Mergler, Pantazopoulos, Schantz and Carpenter2012; Laird et al., Reference Laird, Goncharov, Egeland and Chan2013). Many fish advisories consider multiple chemical contaminants but provide minimal discussion of fish nutrients, focused on omega-3 fatty acids (Scherer et al., Reference Scherer, Tsuchiya, Younglove, Burbacher and Faustman2008). Compared with mercury concentrations, there are fewer studies quantifying fatty acids and selenium in seafood (Karimi et al., Reference Karimi, Fisher and Meliker2014). Therefore, to inform risk assessment more research is needed quantifying the risks and benefits associated with specific seafood consumption habits, such as considering the recommended daily consumption of seafood nutrients relative to reference doses (i.e. hazard quotients) of seafood contaminants (i.e. Gladyshev et al., Reference Gladyshev, Sushchik, Anishchenko, Makhutova, Kalachova and Gribovskaya2009).
To conduct appropriate human health risk assessment for contaminants such as mercury requires an understanding of how mercury, fish oils and selenium co-exposures affect the human body. This work can be informed by studies from marine biology and fisheries science, coupled with epidemiological biomonitoring, anthropological and food science investigations into the role of culinary preparation and gut processing on mercury and nutrient bioavailability (Laird et al., Reference Laird, Shade, Gantner, Chan and Siciliano2009; Moses et al., Reference Moses, Whiting, Bratton, Taylor and O'Hara2009a, Reference Moses, Whiting, Muir, Wang and O'Harab; Costa et al., Reference Costa, Afonso, Bandarra, Gueifão, Castanheira, Carvalho, Cardoso and Nunes2013). Acknowledging the concerns about contaminant exposure from seafood and its health benefits, the Joint FAO/WHO Expert Consultation on the Risks and Benefits of Fish Consumption (2010) recommended that government entities ‘Develop, maintain and improve existing databases on specific nutrients and contaminants, particularly methylmercury and dioxins, in fish consumed in their region’ and ‘Develop and evaluate risk management and communication strategies that both minimize risks and maximize benefits from eating fish’ (FAO/WHO, 2010, p. 33). Nevertheless, their general conclusions acknowledge fish as an important food source with clear benefits for reducing heart disease mortality and supporting optimal neurodevelopment in children.
CONCLUSIONS
Our current ability to properly estimate the risks and benefits to humans of seafood consumption are hampered by the common approaches of separately studying either contaminants or nutrients in fish. To date there are few studies in which fish tissue concentrations have been measured for both contaminants and nutrients across a range of species and geographic regions, even for the restricted set of chemicals considered in this review. There is tremendous variability between and within fish species in their mercury, EPA and DHA concentrations, leading to different versions of the ‘fish intake’ exposure across participants in epidemiological studies (Greenland & Robins, Reference Greenland and Robins2009), complicating the interpretation of studies on seafood health implications. Better characterizing the extent of interspecies and intraspecies variation of chemicals in fish may help inform future human exposure studies by allowing for more explicit accounting of measurement error (Spiegelman et al., Reference Strain, McAfee, van Wijngaarden, Thurston, Mulhern, McSorley, Watson, Love, Smith, Yost, Harrington, Shamlaye, Henderson, Myers and Davidson1997; Murad & Freedman, Reference Murad and Freedman2007; Guo et al., Reference Guo, Little and McConnell2012; Pollack et al., Reference Pollack, Perkins, Mumford, Ye and Schisterman2013). Furthermore, statistical methods are improving for epidemiological studies to incorporate source (i.e. seafood) contaminant levels, intake frequencies, toxicokinetic processes and biomarkers for an integrated exposure assessment (Conti et al., Reference Conti, Cortessis, Molitor and Thomas2003; Bartell & Johnson, Reference Bartell and Johnson2011; Tan et al., Reference Tan, Sobus, Chang, Tornero-Velez, Goldsmith, Pleil and Dary2012; Shin et al., Reference Shin, Steenland, Ryan, Vieira and Bartell2014); or to consider complex interactions between multiple seafood contaminants (Lynch et al., Reference Lynch, Huang, Cox, Strain, Myers, Bonham, Shamlaye, Stokes-Riner, Wallace, Duffey, Clarkson and Davidson2011) Thus, additional research on the joint distribution of multiple chemicals in marine foods has potential to contribute directly to future epidemiological investigations. Bringing multiple stakeholders (i.e. fish consumers and marine scientists) together in a trans-disciplinary conversation with health scientists can also help target the science to relevant questions and improve on knowledge translation (Boote et al., Reference Boote, Telford and Cooper2002; Burger et al., Reference Burger, Gochfeld and Fote2013). Future assessments of the risks and benefits of fish consumption will require more detailed understanding of exposures to both fish contaminants and nutrients as well as the environmental and ecological drivers that control their chemical transformations, and flow through marine food webs. The processes affecting composition of marine fish may be altered by climate change impacts including but not limited to ocean warming and ocean acidification (Edwards & Richardson, Reference Edwards and Richardson2004; Halpern et al., Reference Halpern, Walbridge, Selkoe, Kappel, Micheli, D'Agrosa, Bruno, Casey, Ebert, Fox, Fujita, Heinemann, Lenihan, Madin, Perry, Selig, Spalding, Steneck and Watson2008; Kroeker et al., Reference Kroeker, Gambi and Micheli2012); fishing (Micheli et al., Reference Micheli, De Leo, Butner, Martone and Shester2014); emerging joint exposures such as pharmaceuticals and personal care products potentially changing xenobiotic kinetics for some other compounds (Smital et al., Reference Smital, Luckenbach, Sauerborn, Hamdoun, Vega and Epel2004; Epel et al., Reference Epel, Stevenson, MacManus-Spencer, Luckenbach, Hamdoun and Smital2008; Bosnjak et al., Reference Bosnjak, Uhlinger, Heim, Smital, Franekic-Colic, Coale, Epel and Hamdoun2009); and future changes in contaminant sources and inputs (UNEP, 2013). Together, these changes indicate a need for continued research on fish nutrients and contaminants in marine and medical science, as well as ongoing communication between these disciplines.
FINANCIAL SUPPORT
MOG was supported on a training grant from the National Institute for Environmental Health Sciences (T32ES013678-07). RK's contribution was supported by the Gelfond Fund for Mercury Research and Outreach. MIG was supported by project No. 6.1089.214/K of the Siberian Federal University, carried out according to Federal Tasks of Ministry of Education and Science of Russian Federation, and by Russian Federal Tasks of Fundamental Research (project No. 51.1.1). CYC's contribution was supported by NIH grant numbers P42 ES007373 and 1R01ES021950 from the National Institute of Environmental Health Sciences.