DNA metabarcoding reveals the diet of the invasive �sh Oreochromis mossambicus in mangroves of São Tomé Island (Gulf of Guinea).

Invasive species can trigger profound effects on recipient ecosystems, namely through the food web. Despite being recognized as one of the worst invasive species, little is known about the feeding ecology of the Mozambique tilapia (Oreochromis mossambicus). To understand how this invasive species might impact food webs, we applied metabarcoding to analyze its diet’s composition in two mangroves, in the Obô Natural Park in the oceanic island of São Tomé. Given the particular importance of mangroves as �sh nurseries, we speci�cally aimed to determine if this invader might predate on other �sh species. However, tilapias were mostly phytoplanktivorous, and there were few indications of predation on native �sh eggs or larvae. Instead, tilapias may impact low trophic level resources and nutrient availability with the potential to cascade through the food web by means of bottom-up disruption. In addition, we recorded important changes in the taxonomic composition of the diet, linked to locations and life stages, suggesting that its opportunistic feeding associated with its aggressive territorial behavior may result in resource competition with native species with which it has overlapping dietary niches.


Introduction
The Mozambique tilapia, Oreochromis mossambicus (Peters, 1852) is a cichlid sh native to Eastern Africa, being among the worst and most widely distributed invasive sh species (Global Invasive Species Database, 2023) and occurring in at least 92 countries beyond its native range (Froese & Pauly, 2021).Its introduction was mostly driven by high demand for aquaculture and biological control, which combined with its life-history traits has allowed successful establishment in many new areas (Canonico et al., 2005).Early maturation, highly successful parental care, wide tolerance to environmental conditions (e.g., high salinity variations), and opportunistic feeding behavior enable its invasion of highly variable environments, such as estuaries (Félix et al., 2017).Although, tilapias pose serious threats to native aquatic communities worldwide, through predation and through competition for food, habitat and spawning sites (Russell et al., 2012), little is known about the impacts exerted by its omnivore-detritivore feeding behavior.Thus, it is critical to understand the feeding ecology of the Mozambique tilapia in its invasion range, as this may affect the food web and native assemblages.
The presence of O. mossambicus has been recently described for two mangroves in the São Tomé Obô Natural Park (Félix et al. 2017).The importance of mangrove forests as the source of countless ecosystem services is well established (Primavera et al. 2019).However, it is not clear how tilapias might be disrupting the ecosystem's functioning, since there are no studies that evaluate the impact of tilapias in mangroves.In fact, the majority of the invasion biology research have, by far, been mostly focused on plant species (Chen & Ma, 2015).Moreover, regarding the study of aquatic biological invasions, only 3% of all publications between 1972 and 2012 were from African case-studies (Thomsen et al. 2014).In the context of the availability of genetic resources, there is also an imbalance across geographies and ecosystem types, with many species from tropical aquatic ecosystems remain to be sequenced or even described (Chu et al., 2019).
This study aims to evaluate the diet of the Mozambique tilapia in two mangroves in São Tomé Island.
Speci cally, we aim to (1) determine its diet and (2) examine intraspeci c diet variations related to habitat and life stages.To this aim, we applied DNA metabarcoding, using two genetic markers, to investigate the diet of the Mozambique tilapia in two mangroves of São Tomé.Metabarcoding (i.e. the simultaneous identi cation of many taxa within one sample through next-generation sequencing of a DNA barcode) is regarded as a powerful, reliable, high-resolution method that allows for broad taxonomic coverage being particularly useful when dealing with microscopic prey items (de Sousa et al. 2019).The application of this tool will thus enable us to circumvent the di culties of studying an omnivore-detritivore sh and infer about its putative impact on native biodiversity and on ecosystem functioning.

Field collection
In 2020, we collected 50 tilapia specimens in the mangroves of Malanza and Praia das Conchas (Félix et al., 2017) of São Tomé Island (Democratic Republic of São Tomé and Principe, Gulf of Guinea, Central Africa).The main characteristics of both mangroves are summarized in Félix et al. (2017).Tilapias were captured using cast nets (mesh size 30 mm) and anesthetized using clove oil.All specimens were measured for body mass (± 1 g) and total length (± 1 mm), and then separated into adults or juveniles based on total length (threshold size: 140 mm; Froese & Pauly, 2021).The digestive system of all specimens was extracted through dissection and preserved in 96% alcohol until further processing.

Laboratory processing
DNA extractions of gut content samples were performed using the E.Z.N.A. Tissue DNA Kit (Omega Bio-Tek, Norcross, GA, USA), following manufacturer instructions.Prior to DNA isolation, stomach contents were homogenized with a tissue lyser.PCR ampli cations of two barcodes were prepared in separate PCR reactions.We targeted the 313 bp mitochondrial cytochrome c oxidase subunit I (COI) region with the primer pair of NexF_m1COIintF and NexR_jgHCO2198 (Leray et al., 2013) due to their suitability across a wide range of metazoan species.To identify other eukaryotic prey items, we used the 528F/706R primer set which ampli es a 350 bp region of the 18S ribosomal RNA gene (Ho et al., 2017).Illumina library preparation and sequencing were performed at CIBIO-InBIO's Centre for Molecular Analysis (for details see Appendix 1).

Bioinformatic Analysis
The sequencing data from each barcode was processed separately, and it involved (1) trimming adapter and primer regions and exclude short sequences (< 80 bp) using cutadapt (Martin, 2011); (2) merging pair-ended reads and de ning Amplicon Sequence Variants (ASVs), using dada2 (Callahan et al., 2016); (3) performing blastn runs (Camacho et al., 2009) against the nucleotide NCBI GenBank database (https://www.ncbi.nlm.nih.gov), as well as the SILVA (https://www.arb-silva.de/) and the BOLD (https://boldsystems.org/) databases for the 18S and COI, respectively.A detailed description of taxonomy assignment and exclusion criteria to account for sequencing artifacts, secondary predation and contamination (including from the host's DNA) can be found in Appendix 1.

Diet Analysis
A total of 40 samples passed the quality control steps described above (38 and 36 samples for 18S and COI barcodes, respectively; Table S1).We merged the two datasets by collapsing the taxonomic assignment of ASVs down to the lowest identi able common taxonomic level (da Silva et al., 2019).The combined read count data from 40 individuals was transformed into binary occurrence data, from which the weighted Percentage Of Occurrence (wPOO) was calculated (Deagle et al., 2019).The wPOO simply scales the weight of an occurrence according to the number of food items in the sample.This is a way to account for the relative abundance of each prey per individual without calculating it directly, which is something metabarcoding is known to fail (Deagle et al., 2019).Non-prey taxa such as Fungi, Parasites and Protists were excluded.We arranged the remaining ASVs into 11 broad taxonomic categories, based on their share of the whole diet composition and taxonomic a nities.The category "Other prey" comprised rare taxa (wPOO < 1.6%).We calculated the average wPOO of each food category and compared them between locations and life stages.To this aim, data were square root transformed to build a Bray-Curtis distance matrix, which served as the input to the non-metric multidimensional scaling (nMDS) with 40% standard ellipses representing the core population diet niche (de Santis and Volta, 2021) of tilapias occurring at different sites and different size classes.Differences between groups were assessed using Analysis of Similarity (ANOSIM), following the protocol described by Gkenas et al., (2021).Similarity Percentage (SIMPER) analysis was used to examine the prey categories with the highest contribution to diet dissimilarity, following Gkenas et al. (2021).Diet diversity at the population level was determined by the Shannon-Wiener index (H') using 95% bootstrapped con dence intervals (Gkenas et al., 2021).All analyses were conducted in R (v4.2.1, R Development Core Team, 2022).

Results and Discussion
Out of the reads of acceptable quality, we were able to assign a taxonomic identi cation to 99.9% of the reads for 18S and to 94.5% for COI.We identi ed 102 ASVs assigned to 72 taxonomic groups (17 species, 23 genera, 16 families, and 16 higher-than-family level; Table S2) for 18S and 143 ASVs matched to 33 taxa (8 species, 4 genera, 3 families, and 18 higher-than-family level; Table S3) for COI.Only 56% of the taxa identi ed using 18S and 36% of the taxa identi ed using COI were resolved down to genus.By merging the two datasets down to the lowest identi able common taxonomic level and subsequently removing non-food taxa, we obtained 54 taxa for downstream analyses.The taxonomic resolution ranged from species to phylum, with most taxa resolved down to family level (5 species, 2 genera, 26 families, 1 suborder, 4 orders, 1 subclass, 7 classes, 8 phyla; Table S4).The lower than ideal taxonomic resolution resulted from a lack of a well curated and comprehensive reference database speci c to the study sites.
There was some variation in use of prey between juvenile and adult tilapias, being most notably in Malanza (Fig. 2).ANOSIM corroborated the existence of a signi cant but moderate overlap between juveniles and adults in Malanza (R = 0.284, p = 0.002) but non-signi cant in Praia das Conchas (R = 0.20, p > 0.05).SIMPER analysis indicated that land plants contributed most to dissimilarity between juveniles and adults, with diatoms and green algae being also important for the differentiation in Malanza.Dino agellates had the highest contribution in the comparison between juveniles and adults in Praia das Conchas, with other contributors including rotifers and mollusks (Table 1).Diet breadth was considerable higher for juveniles than adults in Malanza, but patterns remained similar between these groups in Praia das Conchas (Figure S1).The differences in feeding behavior between life-stage groups in Malanza are most probably driven by competition avoidance between the juveniles and adults whereas in Praia das Conchas tilapias are constrained to a smaller area that lacks habitat heterogeneity.One of the few studies on the diet of O. mossambicus in its native range found a shift from a carnivorous to a phytoplanktivorous-detritivorous diet as tilapias mature from juveniles to adults (de Moor, et al., 1986).
We did not observe such trend as phytoplankton was the major component in the diet of both juveniles and adults.The occurrence of a sh prey (Anguilliformes) was only detected once, in a juvenile specimen (Table S5), and given the low prevalence, our results do not corroborate previous studies (Canonico et al., 2005) on the possibility of signi cant predation from tilapia on sh eggs, larvae and juveniles.Nevertheless, this rst study examined a limited number of specimens, in a restricted time period, and considering the key role of these mangroves as nurseries to a high number of species, including commercially valuable ones such as Megalops atlanticus (Félix et al. 2017;Dias, 2022), it would be advisable to perform additional studies.Namely, it would be important to explore the effect of seasonality, which affects mangrove connectivity and regulates sh reproduction seasons.Furthermore, the planktivorous diet of tilapias can also affect the growth and survival of larval sh via habitat changes, competition for shared prey, and resource depletion.It has been demonstrated that the presence of invasive planktivores can dramatically reduce the growth rate and delay ontogenetic habitat shifts during the early life stage of a native sh species (Fletcher et al., 2019).Moreover, the competition for food resources associated with the tilapia's aggressive behavior puts them in a favorable feeding dominance.Gkenas et al., (2022) observed that a non-native cichlid was an extremely effective competitor in face of limited resources against two different species.Finally, since the methodology we employed does not allow differentiating prey from host's DNA, or we did not assess cannibalism, which is documented for this species (de Moor et al., 1986).
The current work constitutes the rst analysis of the diet of the Mozambique tilapia in particularly important habitats, such as mangroves, representing one of the rst attempts to assess the feeding ecology of their invasive populations through the application of an innovative technology.Through the use of DNA metabarcoding we demonstrated that tilapias are mostly planktivorous but can adjust their diet to different environments.Similar opportunistic feeding behaviors has been observed in other introduced cichlid populations and associated to their success in highly variable environments (e.g., Ribeiro et al., 2007).This also calls for further studies to gain a deeper understanding of the impact of tilapias on ecosystems.The conservation of these mangroves ensures the long-term delivery of its ecosystem services, which is of the utmost importance for local human populations.Building further knowledge and disseminating empirical evidence on the impacts of O. mossambicus amongst stakeholders, such as conservation managers, governmental agencies and communities living near impacted ecosystems is vital for the control and prevention of further spread of this invader.
Weighted percentage of occurrence (wPOO) of prey categories detected in the gut content of Oreochromis mossambicuscollected in the Malanza and Praia das Conchas mangroves, in São Tomé, using metabarcoding markers.For more details on taxa categorization, see Tables S4 and S5.

Table 1
Results of the ANOSIM (R) and SIMPER (AvD -Average Dissimilarity) analyses of prey categories in the diet of adult and juvenile Oreochromis mossambicus in two mangroves of São Tomé.Asterisk represents signi cant differences (p < 0.05) in pairwise comparisons.Ranks of prey contributions are shown in brackets.Prey codes are: LDPL, Land plants; DTM, Diatoms; GRLG, Green algae; RTF, Rotifers; DFL, Dino agellates; MLS, Mollusks.