Introduction

In the last few decades, omega-3 long-chain polyunsaturated fatty acids (LC-PUFAs), namely, eicosapentaenoic acid (EPA, 20:5n-3) and docosahexaenoic acid (DHA, 22:6n-3), were the focus of many biochemical, physiological (e.g., Lauritzen et al. 2001; Wall et al. 2010; De Caterina 2011), ecological (Arts et al. 2001; Parrish 2009; Gladyshev et al. 2013; Hixson et al. 2015; Twining et al. 2016), aquacultural (Sargent et al. 1999; Tocher 2015) and nutritional (Simopoulos 2000; Robert 2006; Woods and Fearon 2009; Rubio-Rodriguez et al. 2010; Kouba and Mourot 2011) reviews. These two LC-PUFAs are essential for various physiological and biochemical processes in all vertebrate organisms, including fish and humans. EPA is a precursor in the synthesis of the following bioactive lipid mediators (local hormones)/n-3 eicosanoids: (1) series-3 thromboxanes, which are vasodilators and inhibitors of platelet aggregation and thereby reduce blood pressure; (2) series-3 prostaglandins, which provide anti-inflammatory effects; and (3) series-5 leukotrienes, which reduce allergy symptoms (Broughton et al. 1997; Lauritzen et al. 2001; Kris-Etherton et al. 2002; SanGiovanni and Chew 2005; Wall et al. 2010). In general, n-3 eicosanoids act as counterregulators of n-6 eicosanoids [synthesized from arachidonic acid (ARA, 20:4n-6)], which have opposite metabolic properties to those derived from the n-3 fatty acid, EPA. In turn, DHA is the major structural lipid of retinal, neural and brain cell membranes, comprising 10–30% of total fatty acids (SanGiovanni and Chew 2005; McNamara and Carlson 2006). Moreover, DHA can decrease production of proinflammatory n-6 eicosanoids by inhibiting a key enzyme, cyclooxygenase (Adkins and Kelley 2010; Norris and Dennis 2012).

Studies of many fish species imply a key role of DHA in their neural development and functioning of brain and eye (Sargent et al. 1999; Tocher 2003). EPA in fish, like in other vertebrates, has critical metabolic functions via eicosanoid production, maintaining cardiovascular health, immune and inflammatory responses, and gene expression (Tocher 2015). Reduction of these LC-PUFAs in fish diet affects behaviour, decreases growth rate, development, survival and fecundity, delays response to visual stimuli, decreases burst and cruise swimming speed (Masuda et al. 1999; Francis et al. 2006; Benitez-Santana et al. 2007; Rinchard et al. 2007; Kjørsvik et al. 2009; Vizcaino-Ochoa et al. 2010; Zakeri et al. 2011; Fuiman and Perez 2015; Mozanzadeh et al. 2015).

As mentioned, EPA and DHA are essential for human health. Indeed, for over 30 years, epidemiological studies and clinical trials, including several hundred thousand individuals, indicated that EPA and DHA supplementation considerably reduced the risk of morbidity and mortality of many cardiovascular diseases (Garg et al. 2006; Plourde and Cunnane 2007; Casula et al. 2013). Possible mechanisms by which EPA and DHA improved cardiovascular health included antithrombotic, anti-inflammatory and antiarrhythmic actions (Adkins and Kelley 2010; Phang et al. 2011). The World Health Organization as well as numerous national health organizations recommended personal consumption of 0.5–1.0 g of EPA + DHA per day to reduce the risk of cardiovascular diseases (Harris et al. 2009; Kris-Etherton et al. 2009; Adkins and Kelley 2010; Nagasaka et al. 2014). Furthermore, a daily intake of ~ 1 g of DHA has been recommended to prevent neuropsychiatric disorders and to maintain optimal cognitive function throughout one’s lifespan (Reis and Hibbeln 2006; Robert 2006; Plourde and Cunnane 2007; Dyall 2015; Weiser et al. 2016).

The main dietary source of EPA and DHA for humans is fish (Robert 2006; Adkins and Kelley 2010; Gladyshev et al. 2013, 2015a). Nevertheless, contents of EPA and DHA in edible biomass (muscle tissue) of diverse fish species vary by more than two orders of magnitude (Gladyshev et al. 2013). Therefore, it is difficult to consume the recommended daily intake by eating certain fish species (Kwetegyeka et al. 2008; Vasconi et al. 2015).

For applied science, continual database improvement for EPA and DHA contents in diverse fish species is necessary for an accurate assessment of the intake of these essential nutrients (Harris et al. 2009). Evidently, individuals as well as public health officials should be aware that not all fish are equally valuable sources of EPA and DHA (Chuang et al. 2012). However, there is an acute problem, pointed out in recent reviews (e.g., Hixson et al. 2015): in most published works, EPA and DHA in fish were measured and presented as relative units, namely, percent of total fatty acids. Meanwhile, it was demonstrated that to estimate nutritive value for humans, measurements of LC-PUFAs should be reported per mass of consumed food, mg g−1 wet mass, rather than their percent in total fatty acids provided (Gladyshev et al. 2007, 2012b, 2017a, b; Huynh and Kitts 2009; Woods and Fearon 2009). Therefore, only data regarding fatty acid content on a mass basis, as mg per g of biomass, are suitable for fish nutritive value databases (Litzow et al. 2006).

To understand how EPA and DHA are trophically conveyed, it is necessary to reveal mechanisms that account for the 200-fold difference in EPA and DHA contents in diverse wild fish species. Causes of fatty acid (FA) composition and content variations in wild fish, including those of EPA and DHA, are not completely understood yet (Gribble et al. 2016). There are two groups of factors, that may determine FA fish content: ecological and phylogenetic (e.g., Vasconi et al. 2015). Relative contributions of ecological versus taxonomic factors to FA profiles were quantified for phytoplankton (Galloway and Winder 2015) and diverse marine and terrestrial organisms (Colombo et al. 2017).

Among ecological factors, food has often been regarded as the main determinant of fish FA profiles, especially in aquaculture (Morton et al. 2014; Wijekoon et al. 2014; Betancor et al. 2015). In natural water bodies, ecosystem trophic status (e.g., oligotrophic vs. eutrophic), which resulted in a different quality of phytoplankton as the base of the food web, feeding habits and fish trophic level were reported to determine FA composition via the quality of food resources (Ahlgren et al. 1996; Czesny et al. 2011; Vasconi et al. 2015). For example, the highest values of EPA and DHA were believed to be characteristic of either planktivorous fish or top predators (Tacon and Metian 2013; Hixson et al. 2015; Vasconi et al. 2015). However, other authors reported that PUFA and other FA fish profiles were of genetic character (i.e., species-specific) and could be decoupled from their diets (Sushchik et al. 2006; Kwetegyeka et al. 2008; Gladyshev et al. 2012b; Lau et al. 2012). In addition, it was hypothesized (Ahlgren et al. 2009) that the quality of food resources was the main mechanism controlling PUFA content in herbivorous and omnivorous fish, while for carnivorous fish, the phylogenetic factor (species identity) was more important. Moreover, fish habitat may be important for PUFA contents. For example, marine fish were commonly regarded as having higher levels of EPA and DHA (Garg et al. 2006; Rubio-Rodriguez et al. 2010; Guler et al. 2011). However, there were no statistical comparisons between diverse marine and freshwater wild fish in the available literature (but see Moths et al. (2013) for the sum of omega-3 PUFA percentage). Fish size and swimming speed related to habitat (e.g., pelagic high-mobility fish vs. demersal low-mobility fish) have also been reported to affect EPA and DHA contents (Ahlgren et al. 1996; Tacon and Metian 2013; Vasconi et al. 2015).

Temperature has also been regarded as an important ecological factor determining EPA and DHA contents in fish (Arts et al. 2012). However, there was a discrepancy between results of several experimental studies as well as results of field studies, which should be further investigated (Gribble et al. 2016).

The aim of the present work was to conduct a meta-analysis of our data and published data regarding EPA and DHA contents in various wild fish species. Specifically, we aimed to determine a relative contribution of the following factors to the LC-PUFA content: (1) phylogeny (order identity); (2) type of feeding (trophic level); (3) habitat (marine–freshwater, cold–warm); and (4) size and movement.

Methods

Data

Fatty acid data from diverse wild fish were primarily collected from peer-reviewed, scientific literature. No data on fish reared in aquaculture were included because many variables, used in multidimensional analysis (see below) of ecological features of wild fish, including feeding mode (piscivorous, benthivorous, etc.), habitats (pelagic, demersal, migratory, etc.) and swimming velocity are evidently senseless for fish reared in cages using artificial (formulated) food. We produced two data sets. The first set included data on EPA and DHA contents, mg g−1 of wet mass (WM), from publications in Web of Science, Core Collection on 22 April 2016 for ‘fatty acid AND content AND fish’. From these publications, we selected only those that measured fatty acid content using an internal standard during gas chromatography, and we discarded data that were recalculated from lipid weighing. Data from studies, where wild and cultivated fishes were compared, were screened only for wild specimens (Amira et al. 2010; Heissenberger et al. 2010). We additionally screened studies to select FA data from the white muscle of fish, primarily because this tissue is most often used as the edible portion. Data from other tissues were discarded.

In some of the included literature sources, representative size of the analyzed fish was not reported, but in all cases, fish were obtained from the commercial catch and were adults. Evidently, fish fatty acid composition and content can change during growth and reproductive periods (Faleiro and Narciso 2010; Gladyshev et al. 2010; Fuiman and Perez 2015; Murzina et al. 2016). However, the aim of our present study was to evaluate the nutritive value of commercially caught fish for humans because humans mainly consume wild fish from commercial catches, i.e., fish of representative size. Therefore, we did not take variations of FA content in fish during growth into consideration but instead focused on FA content in fish of representative size.

In various of the articles examined, data on FAs were given relative to fish dry mass (Ahlgren et al. 1994; Heissenberger et al. 2010; Wagner et al. 2010). These data were re-calculated per WM, using data on either water content for individual species (Ahlgren et al. 1994; Chuang et al. 2012) or mean water content values for the relevant order (Gladyshev et al. 2006, 2007; Sushchik et al. 2006, 2007). As mentioned above, the applied aim of the inventory of EPA and DHA contents in diverse fish species is the assessment of their nutritive value for humans, i.e., quantity of their healthy daily personal intake (portion), which is calculated per wet mass (e.g., Kwetegyeka et al. 2008; Chuang et al. 2012; Gladyshev et al. 2013). EPA and DHA content data from Zhang et al. (2012), Neff et al. (2014a, b) and Vasconi et al. (2015) were calculated using the data determined by percentages and total FA contents (mg g−1 WM) specified in these papers. Three evidently artifact values from Cladis et al. (2014) and one from Chuang et al. (2012) were not included in the data set because they severely contradicted all known data. Our unpublished data, included in the set, were obtained using internal standards by methods described elsewhere (Sushchik et al. 2006; Gladyshev et al. 2014). To provide an equal statistical mass of each species reported by several authors on the same species, the mean value of each species was acquired for meta-analysis. There were 172 species in the first set.

The second dataset represented a subset of the first one and encompassed species for which percentage of total EPA and DHA contents and total FA contents (mg g−1 tissue) were additionally reported. There were 88 species in the second set.

Data on fish of representative size, habitat, feeding mode and cruise swimming velocity were acquired from relevant references (Nikolsky 1971; Aleyev 1976; Pavlov 1979; Reshetnikov 2003; Gritsenko et al. 2006; Kukhorenko and Kukuev 2010) or from Internet sources (http://www.fishbase.org/; http://www.fao.org/; http://www.iucnredlist.org/). Regarding cruise swimming velocity (V, m s−1), fish were subdivided into three groups: slow (V < 1), medium (1 ≤ V ≥ 2) and fast (V > 2). This parameter was determined on the basis of data regarding direct experimental estimations, analogies with phylogenetically and ecologically allied species, analysis of the shape of the body and structure of fins (obtained from above cited publications) and using the results of theoretical and experimental studies of swimming of fish and dolphins (Romanenko 2002).

Statistical analyses

To relate variance in fish EPA and DHA content to phylogenetic (species identity) and ecological effects, a redundancy analysis was used similar to Lau et al.’s study (2012). In brief, a gradient length was computed by a de-trended correspondence analysis using the fatty acid data matrix, and length values were 0.869 and 0.733 for the first and second axes, respectively, suggesting linear model responses to explanatory variables (Jongman et al. 1987; ter Braak and Prentice 1988). Therefore, partial redundancy analysis (pRDA) was used for the calculations (Borcard et al. 1992; Legendre and Legendre 1998).

pRDA was conducted with the Vegan package (version 2.4-0) in R (http://cran.r-project.org/). We used the first data set for this analysis, and EPA and DHA contents (mg g−1 of WM) and their sum (EPA + DHA) were used as the response variables. To reduce value distribution skewness, values were ln + 1 transformed. Explanatory variables were grouped into three sets (matrices): phylogenetic, ecomorphological and abiotic environmental features. The phylogenetic set represented the identity of 16 orders (according to Nelson 2006). The ecomorphological set included type of feeding (piscivorous, omnivorous, planktivorous and benthivorous), habitats (pelagic, demersal, benthopelagic and migratory), swimming velocity and size. The abiotic environmental set encompassed two factors: temperature and salinity. Taxonomic orders, type of feeding and habitats were independent nominal data and were coded as a dummy variable (Jongman et al. 1987; ter Braak and Prentice 1988). Other explanatory factors except size (ln + 1 transformed) were used in the RDA as rank-ordered data. First, we applied the redundancy analysis (RDA) for each explanatory matrix, and assessed the global significance using the anova.cca function with 1000 permutations. Then, we conducted a forward selection procedure based on an adjusted R 2 value to reduce the number of explanatory variables (both ‘forward.sel’ and ‘ordiR2step’ functions were used and compared). Only significant (P < 0.05) variables were applied for the subsequent variation partitioning analysis (pRDA) based on the ‘varpart’ function. The significance of each testable fraction in variation partitioning analysis was tested using 1000 permutations. We additionally performed a total RDA for all significant variables selected from the explanatory sets. The forward selection procedure was repeated as well.

Standard errors (SE), Kolmogorov–Smirnov one-sample test for normality D K-S, Pearson’s correlation coefficient r, Kruskal–Wallis H test, and one-way ANOVA with Fisher’s LSD post hoc tests were calculated conventionally using STATISTICA software, version 9.0 (Stat Soft Inc., Tulsa, OK, USA).

Results

Sum of EPA and DHA content in the studied fish species, belonging to 16 orders (data set 1, 172 species), varied from 25.6 mg g−1 WM (Sardinops sagax, order Clupeiformes) to 0.12 mg g−1 (Gymnura spp., order Myliobatiformes) (Table 1). Statistical characteristics of EPA + DHA content for the orders, ranged by maximum values, are shown in Fig. 1. Although maximum values differed between many orders, minimum values were very close to each other, except for those of the order Osmeriformes (Fig. 1). However, there were only three species in the order Osmeriformes (Table 1), and therefore their minimum value should be specified in the future. All values for each order had a normal distribution according to the Kolmogorov–Smirnov one-sample test for normality D K-S, except for the order Perciformes.

Table 1 Content of eicosapentaenoic (EPA) and docosahexaenoic (DHA) acids and their sum (mg g−1, wet mass) in various wild fish species, types of habitat (H1: p—pelagic, bp—benthopelagic, d—demersal; H2: c—cold waters, t—temperate waters, w—warm waters; H3: m—marine, f—freshwater, b—brackish, a—migratory (anadromous)), types of feeding (F: p—piscivorous, o—omnivorous, pl—planktivorous, b—benthivorous, d—detritivorous, ph—consuming phytoplankton, wp—consuming water plants, pph—consuming periphyton), V—swimming velocity: f—fast, m—medium, s—slow, and common or sampled size (cm). Species within orders are ranged by EPA + DHA content values
Full size table
Fig. 1
figure 1

Contents of the sum of eicosapentaenoic acid (EPA) and docosahexaenoic acid (DHA), minimum, maximum and median values and quartiles in fish orders: Clupeiformes (Clu, in parentheses—number of species); Scorpaeniformes (Scorp); Salmoniformes (Salm); Perciformes (Perc); Osmeriformes (Osm); Cypriniformes (Cyp);Siluriformes (Silur); Pleuronectiformes (Pleur); Gadiformes (Gad); and miscellaneous (Miscellan, Anguilliformes, Beloniformes, Beryciformes, Esociformes, Ceratodontiformes, Mugiliformes and Myliobatiformes). Bars—minimum and maximum; boxes—25–75%; black squares—median

Full size image

Analysis of the second dataset (88 species, Fig. 2) revealed an absence of correlation (r = − 0.12, P > 0.05 for log-transformed data) between the sums of EPA + DHA content (mg g−1) and the levels (% of total FAs). Gadus merlangus had the highest percentage of EPA + DHA at 55.8%, while the EPA + DHA content in this species, 0.56 mg g−1, was close to the lowest value (Fig. 2). Using the second data set, correlations between the percentage of EPA and DHA and the content of total FAs (mg g−1 WM) were calculated. There was no correlation between the percentage of EPA and the content of total FAs: r = 0.16, P > 0.05. In contrast, there was a strong significant negative correlation between the percentage of DHA and the content of total FAs: r = − 0.61, P < 0.05 (Fig. 3).

Fig. 2
figure 2

Sum of eicosapentaenoic (EPA) and docosahexaenoic (DHA) fatty acids: levels (% of total fatty acids, grey bars) and contents (mg g−1 wet mass, black bars) in muscle tissue of 88 fish species ranged by levels. (See Table 1 for full fish names)

Full size image
Fig. 3
figure 3

Plot of correlations between DHA percentage and total FA content (mg g−1 wet mass) in 88 fish species: dots—experimental data, line—linear approximation

Full size image

Using RDA based on the forward selection procedure, the significant variables in three sets of explanatory matrices were identified: identity in orders Clupeiformes, Salmoniformes, Scorpaeniformes and Osmeriformes in the taxonomic set, planktivory, swimming velocity and migratory in the ecomorphological set, and temperature and salinity in the set of abiotic environments (Table 2). However, order Osmeriformes, migratory and temperature were excluded from total RDA (P < 0.001) after the forward selection. In addition, variance inflation factors (VIF) were inspected for all remaining explanatory variables, which were low (VIF < 10) and therefore assumed no evidence of collinearity. The highest proportion of explained variation contribution, 16.5%, belonged to the shared contribution of sets of phylogenetic and ecomorphological factors (Fig. 4). The highest proportion of unique contribution, 7.0%, belonged to the set of phylogenetic factors (Fig. 4). In general, all explanatory variables significantly explained 35.6% of the total variance in EPA and DHA contents (Fig. 4).

Table 2 Results from the forward selection procedure, showing significant phylogenetic, ecomorphological and abiotic environmental variables
Full size table
Fig. 4
figure 4

Results of the partial redundancy analysis (pRDA): a the Venn diagram of variation partitioning of EPA, DHA and EPA + DHA fish content, representing unique and shared contributions of three sets of explanatory variables: phylogenetic, ecomorphological and abiotic environments. The significance of each testable fraction was expressed as *P < 0.05, **P < 0.01, ***P < 0.001; b the redundancy analysis (RDA) ordination triplot of the first two canonical axes showing response (EPA, DHA and EPA + DHA) and explanatory variables: arrows and centroids for nominal variables (black triangles). Gray dots represent species. The triplot has 1-st type scaling

Full size image

To specify the above RDA results, the Kruskal–Wallis H test was used because numerous compared variables did not have normal distribution according to the Kolmogorov–Smirnov D K-S test. Concerning the swimming velocity, fast swimming species on average had significantly higher EPA + DHA content than the medium and slow swimming fish (Fig. 5a). For salinity, the most important abiotic environmental factor, migratory (anadromous) and marine species had significantly higher EPA + DHA contents than freshwater-brackish water fish, while freshwater and marine-brackish water species had intermediate values of EPA + DHA contents (Fig. 5b). Regarding type of feeding, planktivorous fish had significantly higher EPA + DHA contents than all other feeding groups, except for omnivores-planktivores (Fig. 5c).

Fig. 5
figure 5

Mean values of the sum of eicosapentaenoic and docosahexaenoic acid content (EPA + DHA, mg g−1 wet mass) in muscle tissue of fish. a Swimming velocity (V, m s−1): fast (V > 2), medium (1 ≤ V ≥ 2) and slow (V < 1). b Salinity of habitats: anadr—migratory (anadromous), fresh—freshwater, fr-br—freshwater-brackish, marine, mar-br—marine-brackish. c Types of feeding: ben—benthivorous, ben/pis—benthivorous/piscivorous (in Table 1, b/p and p/b), pis—piscivorous, omn/pis—omnivorous/piscivorous (in Table 1, p/o, o/p, p/o/pl, o/b/p and pl/p/o), pla—planktivorous, omn/pla—omnivorous/planktivorous (in Table 1, b/pl, pl/b/p, pl/b/wp/d, pl/o, pl/p and pl/ph/p), omn/wp—omnivorous/consuming water plants (in Table 1, b/pl/wp, b/wp, b/wp/d, b/wp/p, d/wp/b, o/b/wp, o/wp, ph/wp/o and wp/b), pla/omn—planktivorous/omnivorous. d Habitat temperature: cold—cold waters, tem—temperate waters, tem/cold—temperate/cold waters, tem/warm—temperate/warm waters, warm—warm waters. e Types of habitat: dem—demersal, dem-pel—demersal-pelagic, pel—pelagic. Bars represent standard errors. Means labeled with the same letter are not significantly different at P < 0.05 after Kruskal–Wallis H test

Full size image

In addition, the remaining explanatory variables, which did not give significant effects in RDA, were analyzed. For habitat temperature, species from temperate-cold waters had significantly higher EPA + DHA contents than species from temperate warm and warm waters, while fish from cold and temperate waters had intermediate values of EPA + DHA content (Fig. 5d). However, pelagic fish appeared to have a significantly higher average EPA + DHA content than that of demersal species (Fig. 5e). It is also worth noting that there were no significant correlations between the representative size of a species and EPA (ln data: r = − 0.185, P > 0.05), DHA (r = 0.03, P > 0.05) and their sum (r = − 0.06, P > 0.05).

Discussion

The highest contents of EPA + DHA were found in fishes which belonged to the order Clupeiformes or Salmoniformes, swam fast, ate zooplankton and inhabited marine waters or migrated from fresh to marine waters (anadromous migrations). Moreover, fish with highest contents of EPA + DHA were pelagic species (naturally, as they were planktivores) and inhabited temperate-cold waters.

Specific traits were associated with high EPA and DHA contents in fish muscle tissues. First, the phylogenetic factor gave comparatively high contribution to the content variations (Fig. 4). The principal role of phylogenetic factors compared to that of ecological factors for FA composition and content was recently demonstrated for aquatic invertebrates (Makhutova et al. 2011; Lau et al. 2012), phytoplankton (Galloway and Winder 2015), birds (Gladyshev et al. 2016) and many marine and terrestrial organisms (Colombo et al. 2017). For fish, phylogenetics also played an important role, as demonstrated in our present study and in the literature (Weber et al. 2016; Colombo et al. 2017). However, according to our data, the interaction of phylogenetic and ecological factors appeared to be of the highest importance for EPA + DHA contents (Fig. 4). Similar conclusion resulted from an in-depth examination of ecological (biome, trophic level) and taxonomic factors (Colombo et al. 2017). Indeed, in the course of biological evolution, a species’ genotype was created by an adaptation to different lifestyles in certain environments. Therefore, high EPA and DHA contents may be regarded as an adaptive feature of fish species. For example, Clupeiformes species mainly inhabit surface waters of open seas and oceans, and therefore they are adapted to fast continuous swimming during long-distance migrations searching plankton productive zones. Fast continuous swimming may be supported by high contents of LC-PUFAs in muscle tissue as follows.

Type of swimming

PUFAs, in particular DHA, were recently proposed to be “pacemakers” for the metabolism of animal cells (Hulbert et al. 2002; Turner et al. 2003; Hulbert 2007). In many vertebrate tissues, including skeletal muscle, a strong positive correlation was found between DHA content of cell membrane phospholipids and rate of metabolism (Hulbert et al. 2002). Polyunsaturated FAs have comparatively low potential barriers for rotation around the carbon–carbon single bonds on either side of the double bonds, and thereby their chains move rapidly, exerting very high lateral pressure on neighboring molecules in a cell membrane (Hulbert 2007). The greater the lateral pressure in the membrane, the greater the activity of membrane-associated enzymes (Hulbert 2007). For example, high DHA content in membrane phospholipids was found to provide higher activity of the ubiquitous enzyme, the sodium pump (Na+, K+-ATPase), which is especially important for providing action potential in excitable cells, including muscle cells or fibers (Turner et al. 2003, 2005; Hulbert 2007). Furthermore, DHA is additionally believed to enhance activity of membrane-bound enzymes of the mitochondrial electron transport chain (ETC); therefore, the most active (high-frequency contraction) muscles that provide high respiration rates have higher concentrations of DHA compared with less active muscles (Infante et al. 2001). Moreover, long-distance migratory birds use high storages of dietary EPA and DHA as performance-enhancing agents to activate membrane-related enzymes of the lipid fuel pathway from adipose tissue to β-oxidation and ETC in muscle mitochondria (Weber 2011). Similarly, the high EPA and DHA contents in muscle tissue of Clupeiformes species appeared to be due to the adaptation for fast continuous swimming. The same may be true for migrating representatives of Salmoniformes.

In addition to providing the metabolic adaptation for fast continuous swimming during migrations, high EPA and DHA contents in a subset of anadromous salmonids, may have one additional ecological cause. Salmonids reproduce in oligotrophic streams and die after spawning. Their carcasses in oligotrophic streams are the main food supply for their juveniles via benthic food chains, which provide valuable food with a high content of n-3 PUFA (Heintz et al. 2004). Therefore, high EPA and DHA contents in the anadromous Salmoniformes may be due to an adaptation for their peculiar way of reproducing in their specific ecological niche while they feed their juveniles with food of high nutritive value that is essential for their growth and development.

Type of feeding

The second characteristic of the fish with high EPA + DHA contents was planktivory, namely, zooplanktivory. Why did planktivorous fish have higher EPA + DHA content compared to that of benthivorous fish (Fig. 5c). The cause may be due to a higher content of EPA and DHA in primary producers, planktonic microalgae, diatoms and dinophytes, compared to that of benthic algae and terrestrial inputs (e.g., Ahlgren et al. 1996, Parrish 2013). Thus, the primary consumers, zooplankton, may have a higher nutritive value for fish concerning EPA + DHA content compared to that of zoobenthos. Moreover, zooplankton mainly consist of comparatively small Crustacea, such as Copepoda and Cladocera, that have thin chitin exoskeletons compared to the hard thick exoskeletons of most benthic invertebrates. Barely digestive chitin exoskeletons evidently compose a different portion of biomass of zooplankton and zoobenthos; therefore, the nutritive value of zooplankton, i.e., content of EPA + DHA per mass unit of organic carbon, appears to be higher than that of zoobenthos. However, this oversimplified presumption cannot be reliably checked at present because most data on LC-PUFAs in aquatic invertebrates were presented in relative units, as percentages of total FAs, while quantitative data, mg of EPA and DHA per g of organic carbon (C), were very sparse. According to these sparse data, species of the dominant taxa of marine zooplankton, calanoid copepod, had an average EPA + DHA content of 17–19 mg g−1 C (Chen et al. 2011, calculated from Table 2 of the reference; Koussoroplis et al. 2011, calculated from Table 2 of the reference). Freshwater zooplankton, composed of Cladocera and Copepoda, had an average content of approximately 19–77 mg g−1 C (Gladyshev et al. 2015b, calculated from Table 3 of the reference). Meanwhile, freshwater zoobenthos (gammarids, insect larvae, oligochaets and gastropods) had an average EPA + DHA content of approximately 22 mg g−1 C (Kalacheva et al. 2013, calculated from Table 1 of the reference). These data generally supported the above presumption regarding the higher nutritive value of zooplankton compared to that of zoobenthos. However, more research should be conducted, especially in marine ecosystems, to compare the nutritive values of zooplankton and zoobenthos regarding LC-PUFAs for elucidating causes of higher EPA and DHA contents in planktivorous fish compared to that of benthivorous fish.

For piscivorous species, their average EPA + DHA content was significantly lower than that of planktivorous species (Fig. 5c). For example, the piscivorous species in the order Clupeiformes, the dorab wolf-herring Chirocentrus dorab, had lower contents of the sum of LC-PUFAs than all other species in this order, which were planktivorous (Table 1). For benthivorous and omnivorous fish, they had an average EPA + DHA content similar to that of piscivorous fish (Fig. 5C). This finding concerning the comparatively low EPA and DHA contents in piscivorous fish contradicted the general belief regarding the increase of these LC-PUFAs with trophic level (Hixson et al. 2015; Strandberg et al. 2015; Colombo et al. 2017). However, to our knowledge, there were no direct quantitative comparisons of EPA and DHA contents as mg g−1 of wet mass or per organic carbon of piscivorous fish and their real prey in specific ecosystems. Therefore, the increased EPA and DHA contents of piscivorous fish require additional research.

It is worth noting that there was no increase in EPA + DHA content in muscles of the arctic grayling, Thymallus arcticus, compared to its food (Sushchik et al. 2006). Therefore, the general impression regarding the increase of LC-PUFA content across trophic levels at present was supported only by data on the trophic pair ‘phytoplankton-zooplankton’ (Gladyshev et al. 2011a, b) rather than by data on fish and their food.

Marine and freshwater environments

According to RDA, the designated abiotic environment ‘salinity’ appeared to be of the lowest importance compared to those of factors from the phylogenetic and ecomorphological sets; however, ‘salinity’ was only moderately important in combination with these two sets (Fig. 4). Indeed, differences between average EPA + DHA contents in marine, anadromous and freshwater species, were not statistically significant (Fig. 5b). What ecomorphological and feeding factors potentially might provide high EPA + DHA contents in marine and anadromous species compared to that of freshwater fish? First, high EPA + DHA contents may be due to fast continuous swimming during long-distance migrations of marine and anadromous species. In freshwater ecosystems, which have small sizes compared to seas and oceans, there is less need and opportunity for long-distance migration. The second factor may be a difference between the nutritive value of marine and freshwater zooplankton. As mentioned above, marine zooplankton consist of copepods, while in many freshwater ecosystems cladocerans are the dominant taxa. Cladocera are known to have significantly lower EPA + DHA contents, mg g−1 C, than Copepoda (Gladyshev et al. 2015b). However, data on EPA and DHA contents in marine zooplankton are too sparse for any relevant quantitative comparison with freshwater zooplankton. This desirable comparison is believed to be possible in the future when relevant measurements are conducted.

In this paper, we focused on the probable advantage of marine pelagic planktivorous species over freshwater species because ranges of EPA + DHA contents of marine and freshwater benthivorous and piscivorous species evidently overlap (Table 1). Meanwhile, fast-swimming marine planktivores, the Clupeiformes, namely, the South American pilchard Sardinops sagax, the longtail shad Hilsa macrura and the European pilchard Sardina pilchardus, had more than a four-fold higher EPA + DHA content than fast swimming freshwater planktivores, the rainbow smelt Osmerus mordax (order Osmeriformes) (Table 1).

In this paper, we regarded marine and freshwater environments as a whole rather than salinity as a separate abiotic variable. Meanwhile, data on the effect of salinity on LC-PUFA percentages in fish were contradictory: both an increase (Xu et al. 2010; Hunt et al. 2011) and a decrease (Cordier et al. 2002; Kheriji et al. 2003) of percentages with an increase of salinity were reported. However, to our knowledge, there were no data on the effect of salinity on EPA and DHA contents (mg g−1 WM) in fish.

Water temperature

In the present work, we did not find any significant differences between average EPA and DHA contents in cold and warm environments, although in temperate-cold habitats, EPA and DHA contents were significantly higher than those in temperate-warm and warm waters. Nevertheless, there is a common impression based on the theory of homeoviscous adaptation that an inverse relationship between temperature and LC-PUFA levels exists. According to this theory, exothermic animals, invertebrates and fish, have a decreased unsaturated fatty acid content with a low melting point in cell membranes and have an increased content of more saturated fatty acids with comparatively high melting points to provide optimal cell membrane fluidity (Farkas et al. 1984; Arts and Kohler 2009). However, many authors questioned a peculiar role of EPA and DHA in the homeoviscous adaptation compared to that of mono-unsaturated and short-chain saturated fatty acids (Stillwell and Wassall 2003; Arts and Kohler 2009; Dymond 2015). Moreover, membrane fluidity or membrane viscosity, in addition to the degree of unsaturation, strongly depends on the type of lipid head-groups as well as the presence of another lipid species, cholesterol (Arts and Kohler 2009; Dymond 2016). Therefore, the notion that the differences in DHA contents between species are dictated by temperature-dependent membrane fluidity needs is simplistic (Infante et al. 2001).

Indeed, literature data regarding the effect of temperature on EPA and DHA contents in fish are ambiguous. In laboratory experiments, some fish species showed an increase of DHA but not EPA under decreased temperature (Arts et al. 2012), while in other species, levels of EPA and DHA remained unchanged when temperature varied (Laurel et al. 2012; Wijekoon et al. 2014). In natural conditions, various researchers found an increase of EPA and DHA in fish from cold waters compared to those from warm waters (Wall et al. 2010; Pethybridge et al. 2015). In contrast, other researchers did not find an increase of these PUFAs in relatively cold habitats and seasons (Gokce et al. 2004; Murzina et al. 2013; Gribble et al. 2016). Evidently, beliefs concerning the simple relationship between water temperature and LC-PUFA contents in fish, which implies higher EPA and DHA production in cold habitats, underestimate the complexity of interactions between the abiotic environment and fish biochemistry (Litzow et al. 2006). Therefore, more work should be completed to determine ecological and phylogenetic mechanisms that control FA composition and content in fish.

It should be emphasized that most data regarding the temperature effect were based on relative measurements, i.e., EPA and DHA percentages in total fatty acids. Meanwhile, the target data for estimation of environmental effects on the nutritive value of fish in humans are based on LC-PUFA contents in the catching biomass. Therefore, in the future, the effect of water temperature should be re-evaluated for EPA and DHA contents.

Percent versus content

There were no significant correlations between EPA + DHA content (mg g−1) and level (% of total FAs). Indeed, there were many fish species with high EPA + DHA contents, ~ 10 mg g−1, and low percentage, < 20%, e.g., chum salmon (Oncorhynchus keta), coho salmon (Oncorhynchus kisutch), lake trout (Salvelinus namaycush), rainbow trout (Oncorhynchus mykiss) and landlocked shad (Alosa fallax lacustris) (Fig. 2). On the other hand, there were many species with a high percentage, > 40%, and low content, < 4 mg g−1, e.g., whiting (Gadus merlangus), Atlantic cod (Gadus morhua), sardine cisco (Coregonus sardinella), Arctic char (Salvelinus alpinus) and humpback whitefish (Coregonus pidschian) (Fig. 2). Meanwhile, there were species with very high contents and percentages of EPA + DHA, e.g., sardine (Sardinops sagax), as well as species with very low contents and percentages, e.g., marbled lungfish (Protopterus aethiopicus) and bonito (Sarda sarda) (Fig. 2).

In contrast, we found a significant negative correlation between DHA content and content of the sum of the total FAs in fish. It is worth noting that the sum of FAs used in our study could be regarded as a proxy for total lipid content because the sum content of total fatty acids in fish was correlated with total lipid content (Ahlgren et al. 1996). A similar phenomenon, namely, a negative relationship of LC-PUFAs to the total lipid content in fish, was reported by other authors (Mairesse et al. 2006, see Litzow et al. 2006). This phenomenon may be explained as follows. EPA and DHA are mostly contained in phospholipids (PL), i.e., in the structural lipids of cell membranes, which should remain constant in proportion to muscle tissues (Mairesse et al. 2006).

In contrast, reserve neutral lipids, triacylglycerols (TAG), which are poor in LC-PUFAs, are of high variance in muscles of diverse fish species (Kiessling et al. 2001; Litzow et al. 2006; Benedito-Palos et al. 2013). Some species, so-called “fatty” fish (Moths et al. 2013), accumulate relatively more neutral lipids that contain predominantly saturated and monounsaturated fatty acids. As a result, the total lipid proportions of EPA and DHA could have become diluted by the accumulation of neutral lipids in muscles, while the LC-PUFA content (as mg g−1 tissue) remained equal compared to that in “lean” fish.

Nevertheless, a relation between contents of total lipids and contents of EPA and DHA there may be more complex. For instance, Kainz et al. (2017) found, total lipid status of fish was better predictor of their PUFA contents, than trophic positions or feeding sources.

Unstudied factors and other uncertainties

In general, all explanatory variables significantly explained 35.6% of the total variance in EPA and DHA contents (Fig. 4). This is a typical variance portion explained in RDA for biological systems (Roy et al. 2014). In similar meta-analysis of FA in phytoplankton, RDA explained 48.4–56.8% of the total variation in phytoplankton fatty acids (Galloway and Winder 2015). However, 64.4% of factors affecting EPA and DHA contents in fish remained unknown. For example, in the freshwater order Cypriniformes the species with the highest content of EPA + DHA was the slow swimming benthivorous Siberian stone loach (Barbatula (= Orthrias) toni) rather than the planktivorous bleak (Alburnus alburnus) (Table 1). In the order Salmoniformes, the planktivorous-omnivorous European whitefish (Coregonus macrophthalmus) with medium swimming speed had approximately a 10-fold higher EPA + DHA content than the planktivorous fast swimming sardine cisco (Coregonus sardinella) (Table 1). Demersal Scorpaeniformes, the sablefish (Anoplopoma fimbria) and the Canary rock fish (Sebastes pinniger), had very high EPA and DHA contents, nearly similar to those of fast swimming pelagic migrants from the orders Clupeiformes and Salmoniformes (Table 1). The marine planktivorous fast swimming Indian mackerel (Rastrelliger kanagurta, order Perciformes) had extremely low EPA and DHA contents (Table 1), which absolutely contradicted the general tendency, found by RDA, except for phylogenetic identity. Therefore, causes of high or low EPA and DHA contents in many fish species still remain unknown and should be explained in future studies.

Unknown factors omitted in our present meta-analysis were eutrophication and pollution. Trophic status of aquatic ecosystems is known to significantly affect EPA and DHA contents in fish. In oligotrophic ecosystems, dominant primary producers, including microalgae, diatoms (Bacillariophyceae), chrysophytes (Chrysophyceae), cryptophytes (Cryptophyceae) and dinoflagellates (Dinophyceae), can synthesize EPA and DHA, whereas in eutrophic waterbodies the dominant taxa are green algae (Chlorophyceae) and cyanobacteria, which cannot produce LC-PUFAs (Ahlgren et al. 1992; Taipale et al. 2016). Therefore, in oligotrophic ecosystems, fish that obtain EPA and DHA from primary producers through trophic chains had a higher EPA and DHA content than those in eutrophic ecosystems (Ahlgren et al. 1996; Taipale et al. 2016). As seen with eutrophication, anthropogenic pollution by organic substances and heavy metals also decreased EPA and DHA contents in fish (Gladyshev et al. 2012a). However, more studies are necessary for quantitative estimations regarding the effects of eutrophication and pollution on LC-PUFA contents in diverse fish species.

Conclusions

The highest contribution of total explained variance for EPA and DHA contents in fish was by the combination of phylogenetic and ecomorphological factors. On average, higher EPA and DHA contents were characteristic of marine planktivorous fast swimming Clupeiformes and anadromous Salmoniformes. Their high EPA and DHA contents were believed to play the role of activators for muscle cell metabolism to support fast continuous swimming, especially during long migrations. Our meta-analysis did not support ideas concerning significant influence of higher trophic levels (piscivory) and cold environments (homeoviscous adaptation) on EPA and DHA contents in fish. There was no correlation between EPA and DHA percentages (% of total FAs) and contents (mg g−1 WM) in fish biomass. Therefore, the meta-analysis confirmed that the percentages were not a reliable measurement to estimate nutritive value of fish species for humans. However, many causes of high and low levels of EPA and DHA in different fish species remained unexplained and should be evaluated in future studies.