The Institute of Marine Research Norway (IMR) conducted mesopelagic trawls in Norwegian fjords (NF, Figure 1 ; including Bjørnafjorden, Boknafjorden, and Osterfjorden) during December 2018 (data published in Alvheim et al., 2020, and Wiech et al., 2020), March 2020 and May 2020, and in the North Sea (NS, Figure 1 ) during March 2020. Two pelagic otter trawls with equal-sized mesh, one 35 m² and one 350 m² openings were deployed ( Table 1 ). AZTI conducted mesopelagic trawls in the Bay of Biscay (BB, Figure 1 ) during September 2019 and September 2020 using a Gloria HOD 352 pelagic trawl ( Table 1 ). After landing, M. muelleri individuals were identified, and a subsample of the catch (approximately 100 individuals in BB and at least 50 in NF and NS) was used to estimate the size distribution of the catch by measuring the standard length. All samples were then frozen on board at -18 °C and kept in the dark to avoid potential degradation of light-sensitive compounds (e.g. vitamins).

From each trawl, we made one or two composite sample(s) by selecting and homogenising at least 25 whole M. muelleri individuals. We analysed important substances (including both nutrients and undesirable substances) based on existing relevant studies (i.e. Alvheim et al., 2020; Wiech et al., 2020) and EU regulations on nutrients and undesirable substances (for all analysed substances, see Appendix 1 ) with analytical methods ( Table 2 ) which are accredited according to NS-EN ISO/IEC 17025 (2017) or are holding the status of National Reference Laboratory (NRL).

We first screened the data. For nutrients with less than 25% of measured values below the limit of quantification (hereafter<LOQ) for all the sampling areas, we chose the upper bound (UB) values for the data analysis, i.e. the LOQ value. Otherwise, we excluded the substance with more than 25% of values<LOQ for at least one sampling area (e.g. vitamin A₂ and vitamin D₃). For all undesirable substances with measured values<LOQ, we used the UB values. The data collected in NF were from three different months ( Table 1 ), they were pooled together due to no statistically significant difference found across the sampling months except for arsenic ( Appendix 2 ).

To decide whether the measured nutrient concentrations could be defined as a source of a given nutrient, we compared them with existing EU regulation values (i.e. thresholds). For minerals and vitamins, we adopted EU Regulation 1169/2011 (https://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri=OJ:L:2011:304:0018:0063:en:PDF). The threshold for a significant source of a given micronutrient refers to whether the product contains 15% or more of the nutrient reference values (NRVs) per 100 g. Here, we calculated these thresholds to concentrations by 15 % × NRV ÷ 100 g . For macronutrients, which only included fatty acids (FAs) in this study, we compared the total ω-3 FAs (expressed as the sum of eicosapentaenoic acid [EPA] and docosahexaenoic acid [DHA]) with the EU Nutrition Claim in EU Regulation 1047/2012 (https://food.ec.europa.eu/safety/labelling-and-nutrition/nutrition-and-health-claims/nutrition-claims_en), which included a threshold for a source of ω-3 FAs, and a threshold that the food item contains high ω-3 FAs.

To decide whether the measured concentration of an undesirable substance was low/high risk to human consumers for a given undesirable substance, we compared them with the maximal levels (MLs) for food established in EU Regulation 1881/2006 (https://faolex.fao.org/docs/pdf/eur68134.pdf).

To categorise the measured substance values from each sampling area, we compared the 25^th and 75^th quantiles with the thresholds and interpreted the categorisation accordingly ( Table 3 ). Because the sample size of substance values for each sampling area was relatively low, we did not assume a normal distribution of the substance values. Thus, to detect differences across sampling areas for each substance, we compared the substance values using the non-parametric Wilcoxon test in Rstudio 1.4.1106 (Team, 2009). The difference between a pair of sampling areas was considered significant if p-value was below 0.05; and such a difference was considered marginal if p-value was not below 0.05 and the median value of one sampling area was outside of the 25^th or 75^th quantile of another sampling area (i.e. the box in the boxplot, Figure 2 ).

There were statistically significant differences in the vitamin A₁ concentrations. The NF and NS samples had higher vitamin A₁ concentrations (1082.86 ± 133.84, and 900.00 ± 167.33 μg/100g [NF, and NS, respectively]; Table 4 , Figure 2 ) than the BB samples (132.50 ± 37.50 μg/100g; Table 4 , Figure 2 ), while no difference was observed in the other micronutrients across the sample areas ( Table 4 , Figure 2 ). Compared with relevant EU regulation values, the M. muelleri from all the sampling areas were shown to be significant sources of selenium/Se (51.00 ± 7.01, 50.10 ± 2.90, 62.10 ± 3.48, and 8.25 μg/100g [BB, NF, NS, and the regulation value, respectively]) and vitamin A₁ (132.50 ± 37.50, 1082.86 ± 133.84, 900.00 ± 167.33, and 120 μg/100g [BB, NF, NS, and the regulation value, respectively]); and the M. muelleri from BB and NS was a significant source of iron/Fe (2.13 ± 0.57, 3.84 ± 1.16, and 2.1 mg/100g [BB, NS, and the regulation value, respectively]; Table 4 , Figure 2 ).

The M. muelleri from NF had higher FA concentrations (e.g. EPA+DHA = 15.98 ± 2.02 mg/g; Table 4 , Figure 2 ) than those from NS (statistically significant; e.g. EPA+DHA = 6.57 ± 0.61 mg/g) and BB (marginal; e.g. EPA+DHA = 10.17 ± 2.20 mg/g). The M. muelleri from all the sampling areas had EPA+DHA concentrations surpassing the relevant EU regulation value for ‘source of ω-3 FAs’ (0.8 mg/g; Table 4 , Figure 2 ).

In wet weight, the M. muelleri from NF had statistically significantly higher concentrations of all selected POPs (e.g. PCDD/F = 0.93 ± 0.13 WHO-TEQ pg/g; Table 4 , Figure 2 ) but lower concentrations of cadmium/Cd (0.03 ± 0.00 mg/kg) than those from BB (e.g. PCDD/F = 0.13 ± 0.02 WHO-TEQ pg/g, and Cd = 0.08 ± 0.01 mg/kg) and NS (e.g. PCDD/F = 0.46 ± 0.08 WHO-TEQ pg/g, and Cd = 0.08 ± 0.01 mg/kg). The samples from NS had higher concentrations of PCDD/F, dioxin+dl-PCB (0.70 ± 0.13 WHO-TEQ pg/g; Table 4 , Figure 2 ), PBDE₇ (0.25 ± 0.04 ng/g), and arsenic/As (4.09 ± 0.19 mg/kg) than those from BB (dioxin+dl-PCB = 0.29 ± 0.06 WHO-TEQ pg/g, PBDE_7 = 0.07 ± 0.02 ng/g, As = 2.25 ± 0.35 mg/kg). The M. muelleri from NF and NS had higher mercury/Hg (0.03 ± 0.00 mg/kg, and 0.02 ± 0.00, respectively; Table 4 , Figure 2 ) and erucic acid concentrations (0.81 ± 0.05 mg/g, and 0.84 ± 0.03 mg/g, respectively) than those from BB (Hg = 0.01 ± 0.00 mg/kg, erucic acid = 0.08 ± 0.03 mg/g). After fat correction, the PCDD/F, dioxin+dl-PCB, and PBDE₇ levels were statistically significantly higher in the samples from NF (e.g. ➤PCDD/F = 0.06 ± 0.01 WHO-TEQ pg/g; Table 4 , Figure 2 ) and NS (e.g. ➤PCDD/F = 0.07 ± 0.01 WHO-TEQ pg/g) than those from BB (e.g. ➤PCDD/F = 0.02 ± 0.00 WHO-TEQ pg/g), while no difference was detected in PCB₆ (1.01 ± 0.38, 0.36 ± 0.02, and 0.38 ± 0.08 ng/g [NF, NS, and BB, respectively]) and PCB₇ levels (1.16 ± 0.44, 0.41 ± 0.02, and 0.42 ± 0.08 ng/g) across the sampling areas ( Table 4 , Figure 2 ).

Considering the established maximum levels (MLs), the M. muelleri from all the sampling areas had PCDD/F, dioxin+dl-PCB (0.29 ± 0.06, 1.73 ± 0.23, and 0.70 ± 0.13 WHO-TEQ pg/g [BB, NF, and NS, respectively]; Table 4 , Figure 2 ), Hg, and lead/Pb concentrations (0.02 ± 0.00, 0.05 ± 0.03, and 0.03 ± 0.01 mg/kg) below the MLs for human consumption ( Table 4 , Figure 2 ). However, the Cd concentrations measured in the BB and NS samples exceeded the ML for human consumption (0.05 mg/kg; Table 4 , Figure 2 ), assuming that the M. muelleri will be consumed whole due to its small size as a convention similar to other small fishes (Roos et al., 2007; Longley et al., 2014; Kolding et al., 2019).

The M. muelleri from all the sampling areas in the North Atlantic Ocean were found to have several benefits. According to relevant EU regulations, they were qualified as significant sources of Se and vitamin A₁ and contained high ω-3 FAs; and those from BB and NS were qualified as a significant source of Fe.

Compared with other mesopelagic species in BB (Chouvelon et al., 2022) and assuming the dry matter content of M. muelleri is 25% (units are in dm, otherwise in ww), the M. muelleri from BB were rather low in the concentrations of Fe (Fe_BB = 85.20, Fe_Others = 47.9-333.3 mg/kg dm), Se (Se_BB = 2.04, Se_Others = 1.68-2.91 mg/kg dm), and Zn (Zn_BB = 51.20, Zn_Others = 24.5-103.7 mg/kg dm). This is the case for the M. muelleri from NF and NS, except that the Se concentration in the NS samples was rather high (Se_NS = 2.48 mg/kg dm).

Compared with other animal food products, the M. muelleri from all the sampling areas had higher concentrations of vitamin A₁ and DHA than whole Sardina pilchardus (vitamin A_1 = 115 ± 32.7 μg/100g and DHA = 0.87 ± 0.15 g/100g; Aakre et al., 2020), a higher concentration of Se than whole Engraulis encrasicolus (Se = 38.2 ± 2.1 μg/100g; Aakre et al., 2020), filet of commercial fishes (e.g. Se _{Salmo salar} = 17 and Se _{Gadus morhua} = 25 μg/100g; Alvheim et al., 2020) and other animal products (e.g. Se_pork/beef = 6 and Se_chicken = 12 μg/100g; Alvheim et al., 2020). These comparisons may involve multiple tissue types, and data from whole fish samples may contain greater variability than that from a single tissue type. Such variability is likely due to different accumulation mechanisms among substances in tissues/organs (Landrier et al., 2012; Shukla et al., 2018), and possibly gut contents of individual whole fish. While this variability may have made the comparisons across tissue types less robust, it nevertheless highlights the possible range in the substance yields from the consumer’s perspective.

Vitamin D₃ is another important micronutrient (Holick, 2007; Amrein et al., 2020) which can be obtained from seafood (Lund, 2013). In our study, most of the measured vitamin D₃ concentrations were<LOQ (LOQ = 0.01 mg/kg) except in two instances (0.01 [from BB] and 0.02 mg/kg [from NS]; Appendix 1 ). According to the EU Regulation No 1169/2011, a food item that provides 15% of 5 μg/100 g vitamin D₃ (or 0.0075 mg/kg in concentration) can be qualified as a significant source of vitamin D₃. Because the LOQ of our method was higher than the regulation value, we cannot confirm whether the M. muelleri collected in this study is a significant source of vitamin D₃.

Malnutrition might be exacerbated due to the pandemic, food shortage (Workie et al., 2020; Zurayk, 2020), and climate change (Gomez-Zavaglia et al., 2020), while the global population is projected to increase. Thus, alternative productive and nutritious foods such as M. muelleri are important to help combat malnutrition, for example, reducing the risk of iron deficiency-induced anaemia (Aikawa et al., 2006; Beck et al., 2014) which is prevalent globally (Stevens et al., 2022). Its high Se concentration may help provide a protective effect against cardiovascular diseases (Mozaffarian, 2009) and Hg poisoning (Gochfeld and Burger, 2021).

The M. muelleri from all the sampling areas were found to have low risk. All the examined undesirable substances had low concentrations except for the high Cd concentrations in the BB and NS samples when compared the measured concentrations against the MLs in relevant EU regulations. We further compared the BB and NS Cd concentrations with the EU regulation on undesirable substances in animal feed under EU Directive 2002/32/EC (https://eur-lex.europa.eu/legal-content/EN/TXT/PDF/?uri=CELEX:02002L0032-20191128&from=EN) and found that they were below the ML (assuming an 88% dry weight). This suggests their potential usage as a feed ingredient, similar to some earlier findings (Berntssen et al., 2021).

The M. muelleri from our sampling areas had lower Cd concentrations than those reported from the Mid Atlantic Ridge (Grimaldo et al., 2020), and lower Hg concentrations than those from the Azores (Monteiro et al., 1996).

Compared with other mesopelagic species from BB (Chouvelon et al., 2022) and assuming the dry matter content of M. muelleri is 25%, the M. muelleri from BB were higher in As concentration than these mesopelagic fishes (As_BB = 90.00, As_Others = 3.83-53.60 mg/kg dm); this is also the case for Cd (Cd_BB = 3.20, Cd_Others = 0.03-2.43 mg/kg dm), and Pb (Pb_BB = 0.80, Pb_Others = 0.03-0.18 mg/kg dm). While the Hg concentrations in the M. muelleri from BB were similar to these mesopelagic species (Hg_BB = 0.40, Hg_Others = 0.090-1.365 mg/kg dm). These differences persist when comparing the M. muelleri from NF and NS with these mesopelagic species except that the Cd concentration in the NF samples was rather low (Cd_NF = 1.2 mg/kg dm), and the Hg concentration in the NF samples was rather high (Hg_NF = 1.2 mg/kg dm).

Compared with other mesopelagic species from NF (Wiech et al., 2020), the M. muelleri from all the sampling areas had lower concentrations of As, and Pb than Meganyctiphanes norvegica (As = 28 ± 19, Pb = 0.086 ± 0.075 mg/kg), As, and Hg than Pasiphaea sp. (As = 22 ± 19, Hg = 0.038 ± 0.02 mg/kg), As, and Pb than Eusergestes arcticus(As = 9.5 ± 4.2, Pb = 0.01 ± 0.006 mg/kg), and Pb than Benthosema glaciale (Pb = 0.016 ± 0.017 mg/kg). The M. muelleri from NF and NS had higher concentrations of total fat and erucic acid than the aforementioned mesopelagic species from NF; while those from BB and NS had lower concentrations of total PCDD/F, PBDE₇, PCB₆, and PCB₇ than these mesopelagic species from NF.

There are other undesirable substances that were not included in the present study. For example, wax esters may pose some health hazards (e.g. keriorrhea) when consumed in large amounts (Schots et al., 2020) and are commonly found among mesopelagic species (Voronin et al., 2022, 2023). However, the wax ester concentrations in the NF and NS samples were low (in both wet weight and % of FA; Appendix 1 ), which is similar to some of the previous findings (Wiech et al., 2020; Voronin et al., 2023).

Both biological and environmental drivers are important in determining the substance concentrations of the M. muelleri in the present study. However, many drivers from these two dimensions are entangled and may synergistically affect the substance concentrations in fish. Apart from the discussed drivers, there are others that may incorporate both dimensions. For example, temporospatial variation in food availability (e.g. algal blooming in NF) can affect the substance concentrations in fish (Boran and Karaçam, 2011) as well as their feeding strategies. Since much of the substances in fish come from their diet (Grimaldo et al., 2020), apart from their geographic distribution (Batool et al., 2016), mesopelagic food webs may be supported by multiple food sources (Ianiri and McCarthy, 2023) with different basal substance compositions (e.g. upwelling, offshore and nearshore phytoplankton, and terrestrial runoff). While the prey specialisation is relatively low among mesopelagic fishes (Bernal et al., 2015), the substance concentrations in fish are thus partially determined by the contributions of prey from different basal food sources. Thus, it is important to understand the nutrient flow in mesopelagic species, which may nevertheless provide insights into the possible environmental impacts and sustainability of potential fisheries. For mesopelagic systems, these are particularly vital (Hidalgo and Browman, 2019; Grimaldo et al., 2020; Standal and Grimaldo, 2021; Fjeld et al., 2023) due to their trophic functions in nutrient cycling, carbon fixation (Li et al., 2022; Schadeberg et al., 2023), and mediating ocean health (Bank et al., 2020).

In this study, we chose nutrients and undesirable substances based on relevant studies and EU regulations. However, there is no EU regulation on a few important undesirable substances including As, wax esters, and PBDE. As they may pose threats to human health, it is important to develop regulations on them to further ensure food safety of current and emerging aquatic foods.

We mentioned the importance of body size on the concentrations of some nutrients and undesirable substances. However, in the present study, only the body size of a group of individuals for each catch was measured, which reduced the resolution in variability to understand the effect of body size on the substance concentrations. It is ideal to analyse the substance concentrations at the individual level. Unfortunately, the current methods require a higher mass of materials than one individual of a small fish such as M. muelleri to analyse the full suite of nutrients and undesirable substances. Thus, future development of analytical methods is essential to enable high-resolution individual-based substance analyses of small mesopelagic fishes.

Nevertheless, we suggested that the season of collection and spawning activities may have caused some uncertainty in the analyses. The sampling of mesopelagic species at IMR and AZTI takes into consideration of the cruise planning and financial feasibility. Targeted sampling of mesopelagic species is challenging, and cruise activity was limited. This led to sampling at different seasons in different areas and varying size distributions. As M. muelleri is known to spawn in batches over a protracted spawning period and shows mixed schooling behaviours, it was not possible to foresee their maturity levels prior to sampling. These factors may have an influence on the substance concentrations. However, our datasets nevertheless have important implications on the geographic differences in the substance concentrations of the M. muelleri in the North Atlantic Ocean, while pointing to important future work to better inform the sampling design and improve the resolution of such analyses by disentangling these potentially confounding factors.