Discussion

Diet profile

The DNA metabarcoding method provided a clear taxonomic resolution of the partially digested stomach content and potentially detected highly diverse food taxa. This high taxonomic resolution is in line with those of previous dietary studies based on DNA-based approaches (e.g., Harms-Tuohy et al. 2016; Pan et al. 2021; Rees et al. 2020; Sakaguchi et al. 2017). Our results indicate that the diet of Campylomormyrusand Gnathonemus species is composed mainly of three types of prey items, i.e., benthic invertebrates, allochthonous invertebrates, and macrophyte material. The most dominant prey taxa found in the gut contents of these species belong to benthic invertebrates, especially aquatic insects. In particular, dipterans (Chironomidae, Simuliidae, Drosophilidae, and Tephritidae), coleopterans (Zopheridae, Carabidae, Histeridae, and Scarabaeidae), trichopterans (Hydropsychidae, Lycaenidae, and Elateridae), ephemeropterans (Leptophlebiidae, Baetidae, and Ephemerellidae), and odonatans (Coenagrionidae, Chlorocyphidae, and Chlorogomphidae) are important diet constituents of all the species. The larvae of these insects live usually in holes and interstitial spaces of the riverbed. Although the DNA approach used here cannot tell the stage of the aquatic insects found in the diet, it is reasonable to assume that the aquatic larvae, rather than the terrestrial imagines, were targeted by Campylomormyrus and Gnathonemus . Indeed, larvae of aquatic insects were reported as food items found in the stomach of mormyrid fish (Blake, 1977; Hyslop, 1986).
Beside aquatic insects, annelid worms (such as Glossoscolecidae, Naididae, and Megascolecidae) were also found in the diet of allCampylomormyrus and Gnathonemus species. Similar to the insects’ larvae, the annelid worms hide in mud and among aquatic vegetation in the substrate of the riverbed. Other benthic invertebrates found were freshwater snails (Gastropoda, orders Pachychilidae and Stylommatophora), and crustaceans (Malacostraca, orders Decapoda, Copepoda, Cladocera, and Amphipoda). Previous studies on the diet of mormyrids using morphological observations did not report annelid worms. The absence of this food taxon may be due to the progressed digestion before the morphological examination.
The second group of food items found in the diet ofCampylomormyrus and Gnathonemus species is allochthonous invertebrates. The most abundant prey taxa from this group are Hymenoptera (including Formicidae, Mymaridae, and Braconidae) and Lepidoptera (including Nymphalidae, Lycaenidae, and Hepialidae). Additionally, Araneae (Arachnida) were frequently found in the diet. These food items are usually located in the habitat at the surface of the water.
The third group of food items is plants, including grasses, such as Poaceae of the Poales order, and flowering plants, such as Fabaceae and Asterales.
It must be noted that we cannot exclude some of these taxa having derived from the diet of the primary prey (secondary predation; Sheppard et al. 2005) or comprise small organisms and plant debris unintentionally ingested during grasp suction. However, the stomach contents of Campylomormyrus and Gnathonemus species found in our study using a DNA metabarcoding approach are compatible withCampylomormyrus mainly (about 90 %) feeding on aquatic insects (Nwani et al., 2008; Roberts & Stewart, 1976). A previous study, based on morphological observation (Roberts & Stewart, 1976), reported that stomach contents of some Campylomormyrus species contain larvae of chironomids, Povilla , trichopteran, ephemeropterans and odonates, dead plant debris, and decomposing animal debris. Roberts & Stewart (1976) reported also that the stomach content of a specimen ofC. rhynchophorus had Chironomidae, Simuliidae, and trichopterans, and a few small ephemeropterans. Another study on the stomach contents of C. tamandua using morphological observation reported similar food taxa (Nwani et al., 2008). The few available dietary studies on other fish species inhabiting the Congo River such as Schilbe intermedius (Dirat et al. 2019) and Distichodus antonii ,D. affinis and D. lusosso (Zebe et al. 2010) showed similar prey spectra.

Diet comparison among Campylomormyrusphenotypes

The dietary compositions among Campylomormyrus andGnathonemus suggest that their feeding behavior is related to phenotypical traits (EOD duration and snout morphology). Based on our results, , different prey spectra were found in the stomach of the different groups of Campylomormyrus species, which may relate to species exploiting diverse food niches in their habitats.
For instance, the prevalence of benthic invertebrates, such as larvae of dipterans and coleopterans and annelid worms, in the diet of all species suggests that they exploit the bottom of the riverbed, while the occurrence of allochthonous invertebrates, such as Formicidae, Nymphalidae and Arachnida spiders, may indicate a certain degree of surface feeding. Further, the diets also include food items from the water column, such as copepods. Accordingly, these species may exhibit high trophic flexibility and diverse feeding behaviors. However, the RRA results revealed significant differences among the dietary compositions among the species, potentially associated with EOD and snout length. For example, the diet of C. compressirostris contains more dipterans, while the diet of C. tshokwe contains more coleopterans.
Species with long EOD had preferentially fed on other taxonomic groups (i.e., coleopterans, ephemeropterans, spiders, annelids), compared to the species with short EOD, where dipterans dominated in the diet. Similar differences were found according to snout length. Note, however, that species with long EOD often also exhibit a long snout (Fig. 4; cf. Lamanna et al. 2016).

Dietary analysis and the radiation scenario

Our study was motivated by potentially providing further support for the hypothesis that radiation of Campylomormyrus is caused by an adaptation to different food sources, associated with diversification of the EOD. The results proved that all the species tested in the current study are able to exploit diverse niches, especially the bottom fauna, regardless of their snout shape and EOD.
The current study still provides some evidence that different snout morphologies and the associated divergence in the EOD translate into different prey spectra. As a different morphology of the feeding apparatus among the species may constitute a functional adaptation to exploiting different substrates (Amen et al. 2020), the different prey spectra could also reflect differential availability of different prey in the respective microhabitats. Our results cannot establish a causal link between EOD characteristics and prey spectra, yet they are compatible with the hypothesis that the divergence in EOD could also be of adaptive value during feeding, beside its proven function as a prezygotic isolation mechanism. In this case, EOD would be a ‘magic trait’ triggering both adaptation and reproductive isolation (Feulner et al. 2009). One approach to further investigate a potential dual function of the EOD divergence (feeding specialization and reproductive isolation) would be to expand on the choice experiments performed by Amen et al. (2020) by presenting a variety of food items. This will though still be confounded by the association of long EOD with long snout. Ultimately, one would like to know prey-specific detection probabilities, relative to the physical properties of the different discharges. To achieve such knowledge remains a challenge.
The presence of distinct prey spectra among Campylomormyrusspecies with different snout morphologies implies a potential association between snout morphology and feeding behavior. Specifically, the species-specific snout morphology in Campylomormyrus may be related to variations in their grasp suction mode of feeding, as previously described by Marrero and Winemiller (1993). Enhanced grasp suction enables them to obtain the aquatic insects that burrow into the different substrate structures. This functional foraging specialization allows for an efficient exploitation of the rich bottom fauna of benthic invertebrates, especially aquatic insect larvae, which other fish may not reach. Such trophic specializations may have triggered the observed radiation of Campylomormyrus in the Congo River.

Potential limitations of our dietary analysis

Sample sizes and quality of DNA. Our study was limited by the availability of stomach content samples we collected from the field, i.e., the Congo River. Despite of a four weeks effort during our 2012 expedition, we did not yield equally high sample sizes for all target species, such that the small number of individuals per species did not allow for diet comparisons among the individual Campylomormyrusspecies, except for C. compressirostris and C. tshokwe . Moreover, all species examined here are nocturnal, so the stomach contents are expected to be highly digested, due to the time between the last meal and the collection during day time. Our samples had been stored for relatively long time (7 years) prior to DNA extraction. We followed established protocols to store the samples (Seutin et al. 1991) which had been successfully applied for degraded samples (Epp et al. 2010; Stoof-Leichsenring et al. 2012). Note also that Campylomormyrus tissue samples taken and stored under the same conditions (i.e., during Congo expeditions in 2004, 2006, and 2012) yielded amplifiable DNA (entire cytochrome b gene, >1000bp) when analyzed up to 10 years later (Lamanna et al. 2016). While we cannot exclude that storage may have contributed to DNA degradation, our positive control (a fresh sample taken from a C. compressirostris raised in our lab and fed with defined food) yielded a similar read number (419,872) as the stomach samples taken in Congo and stored for several years prior to analysis (\(M\text{ean}=335,020\);\(SD=234,220\)). We hence conclude that progressed digestion may be the major cause of the DNA degradation encountered in our study.
Relative Read Abundance (RRA) and biomass correlation. The number of reads belonging to a particular food taxon is utilized as a proxy for the biomass of this taxon to obtain semi-quantitative measures on the relative amounts of different food taxa consumed byCampylomormyrus and Gnathonemus species. This approach is used in most of the dietary studies, and it implicitly assumes a correlation between the number of sequences and the biomass. The logic behind this assumption is that a large biomass of a particular food item in the diet would translate into a large amount of DNA in the dietary sample and, consequently, more recovered DNA sequences (reads). However, potential biases may arise from technical (such as PCR efficiency, primer choice, DNA extraction, etc.) and biological issues (such as among-prey species variation in tissue cell density, survival of DNA during digestion, and the state of digestion), obscuring the quantitative signal. Despite of these caveats, RRA has been considered valid as a proxy for abundance in eDNA (e.g., Kiemel et al. 2022) and sedaDNA (e.g., Krueger et al. 2021) studies. We hence argue that utilizing RRA as a semi-quantitative measure of prey biomass – with all caution – provides relevant additional information relative to merely reporting the presence of prey taxa, which implicitely overstates the importance of rare food items. Moreover, technical biases may affect all samples in a similar manner and hence not compromise our comparison of relative diet composition among different EOD/snout phenotypes.
Taxonomic resolution of dietary items.The dietary differences among the species and phenotypes are more visible in the low taxonomic levels, i.e., at genus and species levels. However, the relatively low percentages of reads assigned to the family, genus, or species level (see Sec. 3.4) may compromise the confidence in the inferred prey spectra at these taxonomic levels. Therefore, we limited the statistical analysis of diet overlap and assessment to the order level (see Sec. 3.4) which we argue could be more reliably inferred (as atested by a lower percentage of unassigned reads). Notwithstanding this limitation, the study confirmed significant differences in the diet among EOD and snout phenotypes. It is plausible to expect that these differences may be even more pronounced at the species/genus levels.