Introduction

Transitional waters are coastal ecosystems originating from both marine and freshwater sources, representing a continuum between continental and marine ecosystems. The peculiar localization of these ecosystems makes them influenced by seasonal, tidal, and freshwater regimes, modifying the ecological niches in terms of salinity gradients, water depth, light availability, and nutrient regimes. This supports complex food webs which contribute to the high biological value of these ecosystems 1,2,3. The existence of such intense gradients makes transitional waters some of the most productive ecosystems, leading to the development of diverse communities and complex self-regulation mechanisms 4,5. Transitional waters provide relevant ecosystem services and societal goods and benefits, some of the most important represented by local mitigation of climate change as well as biodiversity and biomass reservoirs 5,6,7. The important ecological functions of transitional waters depend on the detrital food webs that support carbon and nutrient cycling; for these basic functions, the sediment-associated organisms play a key role 8,9,10.

Key functional groups within sediments belong to benthic macroinvertebrates 11,12,13. Benthic invertebrates are invertebrate animals inhabiting the bottom of water bodies 14. They can inhabit the bedforms surface or live within the sediments, and display different feeding strategies, e.g., deposit-feeding, filter-feeding, grazing, and predation 15. They have a role as suspension feeders by capturing organic particulate from the water column and introducing it to the bottom. Particulate organic material also reaches the sediments by physical deposition and chemical flocculation 16,17,18. Other benthic macroinvertebrates act in the sediment to accelerate the decomposition processes of organic matter 19,20. Benthic macroinvertebrates are suitable as ecological indicators of transitional water ecosystems for their critical role in the food web as well as being sensitive to the biological, chemical and physical conditions of the ecosystems due to their relatively sedentary life, their long life cycles, and their space-use behaviour 21,22,23.

Moreover, the biodiversity reduction of benthic macroinvertebrates can impact the ecosystem functioning as demonstrated for different freshwater ecosystems 24,25. Traditional assessment of benthic macroinvertebrate biodiversity is done through the identification of morphological traits and counting of sampled living individuals. This method requires intense screening work for taxa identification and its accuracy could be undetermined by difficulties in distinguishing phenotypically similar organisms, juvenile stages, and rare species.

In recent years, the application of molecular methods, such as environmental DNA (eDNA) metabarcoding, is improving the analysis of benthic macroinvertebrate biodiversity 26. eDNA metabarcoding refers to the analysis of DNA molecules present in an environmental sample, such as water or sediment, through PCR amplification of specific gene markers, high-throughput sequencing of amplicons, and taxonomic classification of DNA sequences comparing them with DNA barcodes deposited in the reference databases 27,28. DNA metabarcoding can also be applied starting from DNA extracted from a bulk sample of organisms 29,30. The application of eDNA metabarcoding for the biodiversity assessment of benthic macroinvertebrates has been performed in riverine and marine ecosystems starting from sediments 31,32,33,34 and in a few studies from water samples 35,36. In marine ecosystems, the assessment of benthic macroinvertebrate communities through water eDNA metabarcoding has highlighted the need to sample in close vicinity to the bottom and resulted in a poor proxy 37,38.

Multiple databases related to cross-disciplinary research, which allow an in-depth exploration of specialised subfields within a certain academic or scientific discipline, such as Web of Science (http://www.webofknowledge.com/), Science Direct (http://www.sciencedirect.com) and Google Scholar (http://scholar.google.com) have been searched using different combinations of the keywords “environmental DNA”, “metabarcoding”, “benthic invertebrates”, “Mediterranean transitional waters”. Up to date, no works exploring macroinvertebrate communities using eDNA metabarcoding from surface water samples in Mediterranean transitional water ecosystems have been found. Accordingly, for the first time, in this study, benthic macroinvertebrate communities are explored through eDNA metabarcoding from surface water samples in a Mediterranean transitional water ecosystem. The objective of the work is to verify whether the shallow water and limited currents in such ecosystems facilitate the assessment of benthic macroinvertebrate communities through eDNA metabarcoding from surface water samples. Furthermore, we investigated the spatial heterogeneity of benthic macroinvertebrate communities and correlated it to the measured environmental gradients. Finally, we investigated whether the use of biological replicates supports the validity of eDNA metabarcoding as an efficient tool for the assessment of benthic macroinvertebrate community structure in transitional water ecosystems.

Materials and methods

Study area

The study area is represented by the coastal lagoon included within the NATURA 2000 Site “Aquatina di Frigole” (IT50003) located in southeastern Italy along the Salento peninsula in the Apulia Region (Fig. 1), which also hosts the Research Centre for Fisheries and Aquaculture of Aquatina di Frigole of the University of Salento. The Aquatina di Frigole lagoon has a surface area of about 42 hectares and a length of about 2 km. The maximum registered depth is about 1.5 m, and the maximum registered tidal excursion is about 34 cm, annually 39.

Figure 1
figure 1

Map of the Aquatina di Frigole Lagoon showing the localisation of the seven sampling sites. The map was generated using QGIS v3.32.0 (https://qgis.org).

Full size image

The lagoon is directly communicating with the Adriatic Sea through a 400 m long and 15 m wide channel on the southern boundary. On the northern boundary, the Aquatina di Frigole lagoon receives a freshwater input from a lateral ramification of the Giammatteo canal 39.

Sampling and DNA extraction

During the spring of 2021, three replicates of 1 L of surface water were collected in each of the seven sampling sites shown in Fig. 1. The samples were immediately transferred to the Research Centre for Fisheries and Aquaculture of Acquatina di Frigole—University of Salento, Italy, located within the NATURA 2000 site and processed.

To avoid filter clogging and reduce Polymerase Chain Reaction (PCR) inhibitors, a prefiltration process was applied before filtration 40. Specifically, each 1-L replicate has been pre-filtered through a 0.5 mm conical sieve mesh to retain suspended particulate material. Each 1-L replicate was then filtered through a 0.45 μm cellulose filter of 47 mm diameter (Advantec® Mixed Cellulose Ester filters). Following, each filter was singularly used in the DNA extraction process using the DNeasy PowerWater kit (QIAGEN, Germany) following the manufacturer’s protocol. The extracted DNA was immediately stored at − 20 °C.

Abiotic parameters (Salinity, Temperature, pH, and Dissolved Oxygen) are monthly surveyed in the Aquatina Lagoon in correspondence with the 7 sampling sites using the multiparametric probe YSI ProQuatro® (xylem INC, USA).

DNA amplification, sequencing, and bioinformatic analysis

A region of approximately 350 bp of the mitochondrial cytochrome c oxidase subunit I (COI) marker gene was amplified using the degenerated primers BF2 [5′-GCHCCHGAYATRGCHTTYCC-3′] and BR2 [5′-TCDGGRTGNCCRAARAAYCA-3′] 31. The PCR reaction was performed in a volume of 50 μL composed as follows: 5 μL 10X reaction buffer; 1 μL MgCl2 (50 mM); 1 μL dNTP mix (10 mM); 1 μL forward primer (10 mM); 1 μL reverse primer (10 mM); 10 ng DNA; 0.2 μL Platinum Taq (5 U/μL; Life Technologies, USA); sterile water to reach a volume of 50 μL. The amplification program included the following step: denaturation at 95 °C for 5 min, followed by 30 cycles of denaturation (95 °C for 30 s), annealing (50 °C for 30 s), extension (72 °C for 1 min), and a final extension at 72 °C for 10 min.

All PCR products were purified with a PureLink PCR purification kit (Invitrogen, Carlsbad, CA, USA). The Celero™ DNA-Seq Library Preparation Kit cod. 0360A (Tecan, Männedorf, CH) was used for library preparation following the manufacturer’s instructions, without fragmenting the amplicons. Both input and final libraries were quantified by a Qubit 2.0 Fluorometer (Invitrogen, Carlsbad, CA) and quality tested by an Agilent 2100 Bioanalyzer High Sensitivity DNA assay (Agilent Technologies, Santa Clara, CA). Libraries were then sequenced on the MiSeq instrument (Illumina, San Diego, CA) in paired-end 300 bp mode by IGATech, Udine, Italy. High throughput sequencing of the COI amplicons generated a total of 8,654,348 raw reads, with an average of about 412,111 raw reads per sample. An internally developed pipeline and the USEARCH algorithm were used for chimaera filtering, grouping of replicate sequences, and OTU clustering. An internally developed classifier and the IGATech COI reference database were used for taxonomic classification. Specifically, the reference database has been generated by downloading the COI sequences from the NCBI Nucleotide section using the keywords “COI[all fields] OR Cytochrome c oxidase subunit I [all fields]) AND biomol_genomic[PROP] AND animal[All Fields] AND 200:1000[SLEN]”. Steps for the analysis pipeline have been developed as follows: (1) reads were trimmed to eliminate primer sequences by cutadapt1 41, (2) low-quality bases were removed from 3’ by erne-filter2 42 by applying default parameters, excluding reads < 60 bp and (3) reads with an error rate > 1% from further analysis; (4) chimeric sequences were removed with uchime_denovo3 43 implemented in usearch; (5) reads were clustered to a minimum identity of 97% generating representative sequences with cluster_fast 44 implemented in usearch; (6) clusters were blasted against the COI database, without a minimum identity filter; (7) the classifier analysed all the possible blast hits and reported the lowest unambiguous taxonomy possible. The resulting Operational Taxonomic Unit (OTU) annotated list was filtered to contain only macrozoobenthic annotated OTUs.

Statistical OTU diversity estimation

To quantify the macrozoobenthic diversity in terms of OTU richness and the number of reads per OTU in each sampling site, the Shannon–Wiener OTUs Diversity Index has been calculated as a comparative information statistic index. One-way ANOVA was calculated to investigate differences between replicates and between sampling sites. All the above calculations have been done through the software Microsoft Excel. The index results obtained through Microsoft Excel calculations have been uploaded in R and are visualised as a boxplot obtained using the R package ggplot2 v3.5.0 45. To quantify the compositional dissimilarity between sampling sites in terms of the number of reads per OTU, a pairwise Bray–Curtis dissimilarity matrix was calculated for the Wisconsin double standardised data using the R package vegan v2.6-446. The results are visualized as a non-metric multidimensional scaling (nMDS) plot obtained using the R package ggplot2 v3.5.0. Multiple regressions were also calculated to investigate the OTUs which significantly drive the nMDS distribution of the sampling sites and replicates, referred to as intrinsic variables, using the R package vegan v2.6-4. Significance was tested by permutation test (n = 999 permutations). To better understand the differences in terms of OTU composition of the samples and replicates and correlate it to the abiotic characteristics of the study area, a Bray–Curtis distance-based Redundancy Analysis (db-RDA) was performed on the Wisconsin double standardised OTUs data and standardised environmental data using the R package vegan v2.6-4. Response variables are represented by the number of reads per OTU retrieved in each replicate, while explanatory variables are represented by the mean abiotic parameter values registered in the Aquatina Lagoon during the survey held in the same sampling month (Salinity, Temperature, Dissolved Oxygen, and pH). Two-way ANOVA was calculated for significance values. The results are visualised as a db-RDA plot obtained using the R package ggplot2 v3.5.0.

Results

eDNA metabarcoding for the assessment of benthic macroinvertebrate communities composition in the lagoon ecosystem

High throughput sequencing of the COI amplicons generated a total of 8,654,348 raw reads, with an average of about 412,111 raw reads per sample. The observed features and Shannon index rarefaction curve of filtered sequences indicated that the samples reached a plateau (Figs. S1; Fig. S2), highlighting that sequencing depth was appropriate in all the sample replicates. OTUs annotated as macrozoobenthic taxa were selected for further analyses and reported in Table S1.

The total and relative number of OTUs annotated in the seven sampling sites were considered as the sum of the three replicates (Fig. 2). The analysis highlighted a specific and different macrozoobenthic community composition throughout the seven selected sampling sites of the Aquatina Lagoon (Fig. 2, Table S2). Overall, 49% of the OTUs belong to the Mollusca phylum, 16% to the Annelida phylum, 13% to the Gastrotricha phylum, 8% to the Chordata phylum, 7% to the Arthropoda phylum, 3% to the Cnidaria phylum, 2% to the Porifera phylum, and the remaining 2% to the Echinodermata and Nemertea phyla.

Figure 2
figure 2

Total number of OTUs per phylum in each of the seven sampling sites.

Full size image

Analysis of the diversity of the benthic macroinvertebrate communities among lagoon sites

The Shannon–Wiener diversity index (Fig. 3) shows that the sampling sites closest to the seawater inlet (namely AF, A, and B) host a higher macrozoobenthic OTUs diversity with respect to the others. The diversity differences are significant among the sampling sites (one-way ANOVA, p-value < 0.01; df = 6 between groups, 14 within groups; F = 13.96), whereas they are not significant among replicates (one-way ANOVA; p-value > 0.5; df = 2 between groups, 18 within groups; F = 0.01).

Figure 3
figure 3

Boxplot for the Shannon–Wiener Species Diversity Index comparison of macrozoobenthic communities between the Aquatina di Frigole Lagoon sampling sites. Black dots represent confidence intervals.

Full size image

The pairwise Bray Curtis dissimilarity matrix (Fig. S3) and the nMDS ordination (Fig. 4; Fig. S4) clearly suggest a distribution of the benthic macroinvertebrate diversity into three main communities: one localised in correspondence of sampling site AF; one shared between sampling sites A, B, and C; and one shared between sampling sites D, E, and F.

Figure 4
figure 4

Non-metric multidimensional scaling (nMDS), obtained through the Bray–Curtis dissimilarity matrix to quantify the differences in the number of reads per OTU composition of the replicates collected in the Aquatina di Frigole Lagoon.

Full size image

In Table S3, OTUs significantly driving the nMDS ordination (p-values < 0.05) have been marked with an asterisk. In particular, Acroloxus lacustris, Ascidiella aspersa, Botrylloides leachii, Bythinella charpentieri, Bythinella sp., Chaetonotus sp., Distomus variolosus, Euglesa casertana, Gammarus pulex, Gyraulus rossmaessleri, Gyraulus sp., Ichthydium sp., Idiopoma sp., Nais sp., Pseudamnicola sp., Radix sp., Semibalanus balanoides, Squilla mantis, and Theodoxus meridionalis are the species significantly correlated with the nMDS ordination (p-values < 0.05).

Benthic macroinvertebrate communities structure is correlated to the temperature and salinity gradient of the lagoon

To correlate the dissimilarities between the sampling sites and the main abiotic drivers that characterize the environment under study, a constrained ordination has been performed. In particular, a Bray–Curtis distance-based Redundancy Analysis (db-RDA) has been calculated. Four environmental parameters have been used as explanatory variables: Temperature (T), Salinity (PSU), Dissolved Oxygen (DO) and pH (Fig. 5; Table S4). All loading and score values are summarised in Table S5. According to the two-way ANOVA p-values calculated for each of the abiotic parameters, salinity (p-value < 0.01) and temperature (p-value < 0.01) are the main explanatory variables for the OTUs distribution across the sampling sites. pH (p-value < 0.05) and DO (p-value < 0.1) do not affect it.

Figure 5
figure 5

Bray–Curtis distance-based Redundancy Analysis (db-RDA), calculated to explain the differences in the macrozoobenthic taxonomic composition of the Aquatina di Frigole lagoon. PSU = Salinity (Practical Salinity Units, PSU); T = Temperature (°C); DO = Dissolved oxygen (mg/L). Black dots and letters indicate the seven sampling sites and replicates.

Full size image

The db-RDA ordination suggests a clear distinction of the diversity described in the sampling sites related to abiotic factors in different macrozoobenthic communities, showing a distribution similar to that shown in the nMDS ordination.

Discussion

This study represents the first description of Mediterranean transitional waters macroinvertebrate communities' composition and spatial variation using eDNA metabarcoding from surface water samples, replicated three times in each sampling site. Since the lagoon is characterized by shallow water, only surface water was sampled. In addition, sampling surface water avoids using the bottom sediments that are characterised, in these peculiar coastal ecosystems, by a high concentration of humic substances. These substances are natural organic compounds derived from the decomposition of plant and animal material and are important components of lagoon ecosystem sediments47,48,49. The presence of humic substances in natural samples negatively interferes with the extraction of high-quality environmental DNA and with the PCR amplification process 50,51,52. Through the use of surface water samples, we overcame these issues, and we were able to obtain results coherent with the environmental features of transitional water ecosystems. Indeed, we analysed the benthic communities’ spatial variation by collecting surface water samples along the natural gradients of the lagoon, from the connection with the sea to the freshwater input. Specifically, salinity and temperature are two key environmental factors that play a crucial role in shaping the structure and function of lagoon ecosystems 53,54,55. Salinity and temperature in lagoon ecosystems can have profound effects on the distribution and abundance of species 56, as well as the overall productivity and biodiversity of the ecosystem2,23,57,58. High throughput sequencing results from this study demonstrated their reliability in defining the spatial heterogeneity of the study site, with species occurrences distributed according to the monitored salinity and temperature gradients. For instance, despite the high connection and circulation rates generally characterising lagoon ecosystems 59,60,61,62,63, in this eDNA metabarcoding assessment we identified freshwater species like Bythinella charpentieri, Euglesa casertana, Idiopoma sp. and Radix sp. in the lagoon sites closer to the freshwater input. At the same time, we identified marine species like Botrylloides leachii, Distomus variolosus, Semibalanus balanoides and Squilla mantis in the lagoon sites closer to the sea connection. Finally, we identified more broadly distributed species, like the brackish Nais sp. along the entire lagoon.

In this study, eDNA samples were amplified using universal primers targeting a fragment of the mitochondrial COI gene. This region of the mitochondrial DNA has been chosen for its broad records in the reference libraries databases for animal species 64 and for benthic macroinvertebrates in understudied regions 26. In particular, the BF2/BR2 COI primer pair has been selected due to its higher detection and precision rates, compared with other combinations, for benthic macroinvertebrates 65.

The obtained results confirm the validity of this primer pair in detecting a wide range of benthic macroinvertebrate taxa. Considering the overall biodiversity of benthic macroinvertebrates detected via eDNA metabarcoding, we have identified families such as Gammaridae, Hydrobiidae, Muricidae, Palaemonidae, Penaeidae, and Spionidae, common in this ecosystem as reported in the most comprehensive morphological inventory of species of Aquatina di Frigole lagoon66. Not all families reported in the morphological inventory have been detected. Probably, this is due to the absence, concerning less-studied environments, of the publicly available molecular sequences of several species, hindering the detection of their presence 67. The OTUs obtained in this study and unassigned at the species level represent about 40% of total OTUs. This data agrees with the previous gap analysis of COI DNA barcodes in the international databases for transitional waters benthic macroinvertebrate species of the region in which Aquatina di Frigole lagoon is located 67. In this previous study, we reported the absence of a DNA barcode for 36% of benthic macroinvertebrate species. In more detail, both gap analysis and eDNA metabarcoding assessment showed that Mollusca represents the Phylum with the highest number of OTUs unassigned to the species level (Table S1) 67. This suggests the ability of eDNA metabarcoding assessments to recover high biodiversity rates in agreement with the observed gaps in the international and publicly available DNA barcode reference libraries.

Furthermore, the high throughput sequencing data identified 48 families (Table S1) not reported in the morphological inventory66 , including easily misidentified families such as Naididae and Talitridae 68,69. This suggests the role of molecular surveys in increasing the resolution of biodiversity assessments by unveiling the presence of hard-to-identify taxonomic groups.

In conclusion, the results from this study appear robust to state that the eDNA metabarcoding high-throughput data obtained from surface water samples allow the description of the benthic communities’ composition, as well as their spatial variation, in shallow coastal lagoon ecosystems, opening for more rapid and reliable monitoring programs of such valuable ecosystems.