Discussion

The present study confirms, for the Bay of Biscay, the global trend towards observing larger fish at greater depth for deep pelagic fish communities, as described in other systems (Badcock and Merrett, 1976; Willis and Pearcy, 1980; Auster et al., 1992; Stefanescu and Cartes, 1992; Gartner Jr et al., 1997). The presence or not of ontogenetic shifts and their associated main drivers (trophic and/or habitat) were described: i) species undergoing both ontogenetic shifts in vertical habitat use (as inferred from trawling data) and in their trophic ecology (as inferred by δ15N values):L. crocodilus and M. atlanticum ; ii) species showing only an ontogenetic shift in their trophic ecology (i.e. the significant influence of size on δ15N values): M. punctatum, A. risso, S. beanii, S. boa, and A. carbo ; iii) species for which only depth influences their δ15N values (X. copei andA. olfersii ), although X. copei also presented an ontogenetic shift in habitat use form trawling data (with the smallest individuals not found in the deepest stations); and iv) species showing no ontogenetic shift: L. macdonaldi , S. koefoedi, andN. kroyeri .
When studying ontogenetic shifts in habitat use and trophic ecology within deep pelagic fish communities, several important aspects must be considered. The first limitation is the small depth or size range sampled for some species. In particular, A. carbo and S. boa were sampled over a depth range of less than 100m, so relationships between δ15N values and/or size with depth could not be estimated. Alternatively, L. macdonaldi was sampled over a small size range (≈3 cm), which may potentially explain the lack of significant relationships found with individual size for this species. As the isotopic sampling took place over 14 years, a possible temporal bias in the δ15N values cannot be excluded. However, more than 90% of the muscle samples were collected between 2019 and 2021, and almost 75% in 2021, reducing this potential temporal variability.

Community level

A significant increase in the size of individuals with depth was observed at the community level. This pattern has already been observed for migratory species, with older stages generally found at greater depths than younger ones, as individuals that may have reduced their migratory range or stopped migrating with age (Badcock and Merrett, 1976; Clarke and Wagner, 1976; Nafpaktitis, 1977; Willis and Pearcy, 1980; Lancraft et al., 1988; Auster et al., 1992; Stefanescu and Cartes, 1992; Gartner Jr et al., 1997). This general trend in the deep pelagic realm may be a consequence of the trade-off between foraging and predation. To satisfy their energetic needs, juveniles and adults of some species migrate to the epipelagic layer to feed at night. Alternatively, at the senescent stage, some species undergo a reduction in swim bladder size as they age (e.g. L. crocodilus ), so that the energetic cost of migration may be greater than the benefit provided. Some of these species, therefore, adopt a benthopelagic behaviour, which allows them to reduce the energy expended on foraging by taking advantage of the higher concentrations of zooplankton in the benthopelagic zone (Angel and de C. Baker, 1982; Vinogradov, 2005).
At the community level, a slight but significant increase in δ15N values with individual size was also found. However, this relationship was very weak (R² = 0.01). Indeed, it has previously been shown that within fish communities, the increase in δ15N values as a function of individual size was more strongly linked to ontogenetic changes than to the fact that the largest species in the community fed on higher trophic-level prey (Jennings et al., 2002; Stowasser et al., 2012). There are several possible causes: omnivory, large predators feeding on smaller prey, large pelagic suspension feeders feeding on small suspended particulate organic matter, and morphological adaptations of small predators to feed on larger prey (Jacob, 2005; Jennings, 2005; Bode et al., 2007). In particular, in our study, S. beanii was the second largest species sampled (= 55 cm) but had the lowest mean δ15N values (= 9.5‰). This could be partly explained by their serpentine morphology, with a large individual size that is not proportionally reflected in the size of the mouth opening, limiting their ability to capture large prey compared to other species with similar individual sizes. Indeed, our study includes a wide variety of morphology, including lanternfish, hatchetfish, dragonfish, and eelfish. This diversity in shape is bound to lead to large differences in feeding strategies. Consequently, body size may not be the best measure to infer the trophic ecology of these species, although it is often correlated with several other measures. For example, body mass, size, or shape of the mouth opening may be more relevant to study this relationship at the community level (Villéger et al., 2017).

Trophic-driven ontogenetic shift

Intraspecific ontogenetic shift in fish trophic ecology (as inferred from δ15N values variation with size) is generally a consequence of the ability of fish to catch larger prey. This ability is proportional to mouth size, which in turn is proportional to body size, allowing species to feed on larger prey (Dunic and Baum, 2017). Such a pattern has already been observed in mesopelagic fish species (Gartner Jr et al., 1997). In the present study, half of the studied species showed a significant change in their δ15N values with the increasing size of individuals. Of these species, four had their δ15N values influenced solely by the size of individuals: M. punctatum, S. beanii, S. boa, and A. carbo . However, in the case of A. carbo and S. boa , the depth range sampled for these species was too small to test the other relationships. M. punctatum was previously described as a generalist feeder in the Mediterranean Sea with a mixed diet during all stages of its development, except small individuals that seem to feed exclusively on copepods (Scotto di Carlo et al., 1982). As they grow, individuals become more efficient predators and begin to select larger, more nutritious prey (Bernal et al., 2015). In another study in the Northern Atlantic, A. carbo also showed an increase in δ15N values with individual size, confirming the probable ontogenetic diet or trophic level shift for this species (Farias et al., 2014). Its diet would shift from pelagic zooplankton to bathypelagic prey, reflecting an improvement in its predatory ability (Farias, 2014). To our knowledge, this is the first time that an ontogenetic shift in trophic ecology is reported for S. beaniiand S. boa . The diet of S. beanii is generally described as being composed of crustaceans and small fish, whereas that ofS. boa is composed of crustaceans and mesopelagic fish, so our results suggest that the proportions may vary with individual size (Whitehead et al., 1984). Finally, among the studied species, onlyA. risso has undergone a significant but weak decrease in δ15N values with individual size. This trend has already been found for two species of the southern Kerguelen mesopelagic community belonging to the families Platytroctidae and Myctophidae (Woods et al., 2019). A significant decrease in δ15N values was also found in a small pelagic neritic fish species, for example, i.e. the European sardine (Sardina pilchardus ). This reduction was attributed to the greater efficiency of large sardines in capturing phytoplankton, which is less enriched in 15N than zooplankton prey (Bode et al., 2003, 2004, 2006). In addition, several species belonging to the Paralepididae family (which includesA. risso ) have shown tooth loss in adult specimens and a recent study in the western Atlantic also found this pattern for A. risso (Ho and Duhamel, 2019; Devine and Van Guelpen, 2021). This tooth loss may lead to dietary changes in this species, partially explaining the negative relationship between δ15N values and size found in our results.

Trophic and habitat-driven ontogenetic shifts

L. crocodilus and M. atlanticum showed, in addition to an increase in δ15N values with individual size, an increase in individual size with depth. In the Mediterranean Sea,L. crocodilus has already been shown to make a change in its diet in relation to its changes in migratory activity (Stefanescu and Cartes, 1992; Fanelli et al., 2014). L. crocodilus has a diet dominated by epipelagic crustaceans in its pelagic phase and its migratory activity decreases or even stops when it reaches the senescent stage. It then adopts a benthopelagic behaviour and feeds on fish at greater depth (Stefanescu and Cartes, 1992; Valls et al., 2014; Bernal et al., 2023). Our present results, therefore, suggest that this behaviour may also occur in the Bay of Biscay. As for M. atlanticum , it has a particular mode of reproduction. This species adopts a benthic behaviour during the spawning period and the fertilization of its eggs takes place in burrows located under the surface of the sea bed (Silverberg et al., 1987; Dallarés et al., 2021). This specific reproductive behaviour may explain our results that the largest, and thus reproductive individuals are found at greater depth. From a trophic perspective, it has been shown in the Mediterranean Sea that the diet of M. atlanticumconsists almost exclusively of pelagic prey (Dallarés et al., 2021). Larger individuals may have the ability to capture larger prey, which can also explain the relationship between δ15N values and individual size found for this species. These two species (L. crocodilus and M. atlanticum ), in addition to having these two significant relationships (larger individuals are found deeper and have higher δ15N values), presented the highest percentages of the variance in δ15N values explained by both size and depth (i.e., the red portion, >7%) in the partition models. This part reflects the proportion of the model that cannot distinguish the effect of depth and size on δ15N values. The benthopelagic behaviour of these species in the adult stage may also partly explain this influence, as δ15N values (including those of POM at the base of food webs) are higher at greater depth and particularly in the benthic domain (Bergmann et al., 2009; Trueman et al., 2014; Richards et al., 2020).

Depth-driven increase in δ15N values

Two species, X. copei, and A. olfersii had their δ15N values significantly influenced only by depth (and not by individual size). In addition, in the case of X. copei the smallest individuals were not found at greater depth. LikeM. atlanticum, X. copei was previously reported to spawn demersally in the North Atlantic, with individuals in pelagic trawls that were juveniles and larger fish that were caught in the deeper stations near the bottom (Mauchline and Gordon, 1983). As the sampling in our study was carried out during the spawning season of this species (October-November), many spawning individuals were observed. The capture of large spawning individuals of X. copei at depth suggests that spawning of this species also occur on the slope area at that period in the Bay of Biscay. Although X. copei did not show a significant relationship with size, it stood out for its wide dispersion of values δ15N values (CV = 6.57). In the North Atlantic, pelagic individuals of this species have a diet limited mainly to copepods and ostracods, while benthic individuals show a wider variety of food, maybe explaining in part this high variability in δ15N values (Mauchline and Gordon, 1983). In addition, it has been observed that both immature and larger individuals may graze on inactive prey or debris such as small copepods, potentially decoupling the relationship between size and δ15N values (Roe and Badcock, 1984). In addition, the δ15N values of X. copei were significantly influenced by depth. This may be a consequence of the reproductive individuals residing at depth for this period and therefore integrating the enrichment of the δ15N values at depth (Choy et al., 2015; Gloeckler et al., 2018; Romero‐Romero et al., 2019; Richards et al., 2020). As forA. olfersii , it is a short-migratory species described to feed on crustaceans and small fish (Muus et al., 1999). It has already been shown that non-migratory species such as A. olfersii integrate changes in zooplankton δ15N with depth (Koppelmann et al., 2009; Hannides et al., 2013). Indeed, individuals located deeper are more dependent on the food web based on bacterially degraded organic particles and thus enriched in δ15N than individuals located less deep (Choy et al., 2015; Gloeckler et al., 2018; Romero‐Romero et al., 2019; Richards et al., 2020). This result was already observed for two non-migratory species near the island of Hawaii: Cyclothone pallida and Melanocetus johnsonii(Romero‐Romero et al., 2019). Among the non-migratory species in our study, A. olfersii was the species with the largest range of depth sampled (≈1000m) which may explain the significant influence of depth on δ15N values for this species. For the other non-migratory species (i.e. L. macdonaldi, A. risso, S. koefoedi ), the depth range sampled was maybe too small to detect any influence of depth on the δ15N values.

No ontogenetic change and no influence of depth on δ15N values cases

Finally, five species showed no trophic-driven ontogenetic change:X. copei, S. koefoedi, A. olfersii, L. macdonaldi, and N. kroyeri . Among these species, some showed high variability in δ15N values (X. copei, S. koefoedi, A. olfersii ) while others had relatively constant values across their size range (L. macdonaldi andN. kroyeri ). This result could potentially reflect differences in feeding strategies between species. Species with a high dispersion of δ15N values may have higher dietary plasticity, allowing them to feed on a wide variety of prey. Such a pattern has already been found for several small pelagic neritic species such asScomber scombrus in the Iberian Peninsula (Bode et al., 2006). However, information on the diet of the Platytroctidae family is very scarce in the literature. S. koefoedi has been reported to have a diet composed largely of copepods, but also ostracods, chaetognaths, and polychaetes, which could partly explain the large variability in δ15N values found during its ontogeny (Hopkins et al., 1996; Novotny, 2018). In contrast, species with low variability in δ15N values could have implemented an alternative strategy to the one classically observed, based on an increase in the size of the prey associated with an increase in mouth size. In this case, meeting energy requirements would be based on an increase in the quantity of resources ingested, made possible by the increase in mouth size, while maintaining the same type (size) of food resources rather than a higher energy content per larger prey ingested. The two Myctophidae species of this group, N. kroyeri, and L. macdonaldi showed weak variability in δ15N values through their ontogeny. These two species have been reported to have a diet mainly composed of crustaceans (Gjøsæter, 1981; Coad and Reist, 2004). However, for the L. macdonaldi case, the restricted size range sampled can explain part of this absence of relation (= 3 cm). In addition, the smaller individuals sampled had a standard length of 11.5 cm, which is important considering the maximum size of 16 cm reported for this species. In the case of N. kroyeri , individuals’ diets may not be size-restricted like in the case of filter feeders, with a strategy of increasing the quantity of ingested food with size.

The Myctophidae case study

An important observation from our results is that the Myctophidae species studied here appeared to have significant differences in their feeding strategies during ontogeny. They were found in three of the four groups of species formed from the different relationships. First,L. crocodilus differs from the other species in that it undergoes ontogenetic changes in both its diet and depth distribution. As large adults have a reduced swim bladder, the energy gain associated with nocturnal migration to feed in productive surface waters may outweigh the costs, making feeding in the benthic boundary layer more cost-effective (Fanelli et al., 2014). While both M. punctatumand N. kroyeri are known to migrate vertically at night in the epipelagic layer, they appeared to have adopted opposite feeding strategies, with M. punctatum appearing to change its diet during ontogeny (i.e. increasing δ15N values with size), whereas not only was this change not seen in N. kroyeri , but its δ15N values remained very stable across the species size gradient. Thus, by not shifting its diet towards larger or more energetic prey during ontogeny, N. kroyeri appears to have opted for an increase in food quantity rather than a change in quality. Finally, L. macdonaldi had the deepest distribution (i.e. median depth = 2000 m), which probably explains in part the lack of profitability for this species to move into the epipelagic layer to feed at night. L. macdonaldi also appeared not to have undergone any dietary changes during ontogeny, although this remains to be confirmed with a wider sampling across the size range of the species. All these differences in the ontogenetic foraging strategies of these phylogenetically related species may be partially explained by differences in morphological traits. Indeed, differences in morphological traits can lead to differences in swimming ability, prey capture, detection ability visual acuity, etc (Villéger et al., 2017). Differences in feeding strategies and depth distribution during ontogeny suggest divergence within this highly diverse family to avoid competition.

Conclusions

Mesopelagic species are a key element in the pelagic food web, as they provide an intermediate link between zooplankton and top predators (e.g. tuna, marine mammals) (Pauly et al., 1998; Pusineri et al., 2005, 2007; Griffiths et al., 2013). Given the diversity of strategies adopted by mesopelagic species, it seems difficult to predict the presence or not of ontogenetic changes. However, this basic information is essential to build functional trophic models that are relatively faithful to reality. For these reasons, the knowledge of their ontogenetic changes must be strengthened to better estimate the ecosystem services they provide and to assess the environmental risks associated with their exploitation (St. John et al., 2016; Hidalgo and Browman, 2019; Christiansen et al., 2020; Drazen et al., 2020).