Introduction

In recent years, the use of extra-organismal DNA has become a widespread method of monitoring vertebrate organisms in freshwater, brackish water and marine ecosystems (Lynsey R. Harper et al., 2019; Masaki Miya, 2022; Sigsgaard et al., 2020; Wang et al., 2021). Among a compilation of 358 vertebrate studies in aquatic environments using environmental DNA (eDNA) methodologies (Supplementary Data 1), fish were the most targeted group, followed by amphibians, mammals, reptiles, and birds (73.3, 17.4, 5.1, 2.3 and 1.8%, respectively).
When targeting only one species (taxon_specific studies), conventional PCR allows the detection of species (Ficetola, Miaud, Pompanon, & Taberlet, 2008; Jerde, Mahon, Chadderton, & Lodge, 2011), whereas quantitative real-time PCR (qPCR) and droplet digital PCR (ddPCR) are the main eDNA methodologies for increasing the species detection sensitivity and quantifying the abundance of DNA sequences (Olsen, Lewis, Massengill, Dunker, & Wenburg, 2016; Takahara, Minamoto, Yamanaka, Doi, & Kawabata, 2012), enabling the indirect estimation of absolute species abundance (Wilcox et al., 2016; Yates, Glaser, et al., 2021). Species assemblages can be identified by metabarcoding after amplification via PCR of one or more genomic regions provided that the appropriate species reference database is available (Miya et al., 2015; Valentini et al., 2016). The number of reads per species is used as a proxy of the relative abundance of species (Di Muri et al., 2020; Goutte, Molbert, Guerin, Richoux, & Rocher, 2020; Pont et al., 2018). Metabarcoding is less frequently used than taxon-specific studies (117 from a total of 358 publications) and is mainly used for fish (86% of papers) in marine and river ecosystems (36% and 37% of papers, respectively, Supplementary Data 1).
Both taxon-specific and metabarcoding approaches are in general more efficient than traditional sampling methods for detecting species (Czeglédi et al., 2021; Hanfling et al., 2016; McElroy et al., 2020; Pont et al., 2018; Valentini et al., 2016), even if the scale of inference in space and time for an eDNA sample must be better defined (Deiner et al., 2017). Comparisons between taxon-specific and metabarcoding approaches are scarce. Additionally, the taxon-specific method has been reported to be both more robust and sensitive than metabarcoding (Bylemans, Gleeson, Hardy, & Furlan, 2018) and equivalent to metabarcoding (Harper et al., 2018). Depth sequencing, number of technical replicates and occupancy modelling are also key factors that can improve the robustness of metabarcoding (Ficetola et al., 2015; Harper et al., 2019).
The number of eDNA copies in a sample obtained by taxon-specific studies (qPCR) is a significant proxy for both density and biomass (Doi et al., 2015; Takahara et al., 2012; Wilcox et al., 2016) but remains a rough estimate of aquatic vertebrates (Ushio et al., 2018) Ninety percent of a compilation of 63 studies identified significant relationships between eDNA concentrations and the abundance or biomass of target species (Rourke et al., 2021). However, this relationship is generally of medium strength due to the huge numbers of factors affecting the production, degradation, transport, sedimentation, and detectability of eDNA particles in relation to ecological/physiological species characteristics, advection/diffusion processes, temperature, pH or bacterial activities (Deiner et al., 2017; Rourke et al., 2021; Yates, Cristescu, & Derry, 2021). A meta-analysis based on 19 studies (Matthew, Yates, Fraser, & Derry, 2019) showed that the correlation is higher in controlled experiments than in the field (82% and 51% of the total variance explained, respectively), partly due to the uncertainties associated with the field estimation of organism abundance by the conventional sampling method (Di Muri et al., 2020).
Metabarcoding provides only the number of reads per taxon that are not related to the amount of corresponding eDNA extracted from the water sample. The relative number of reads is a good proxy for the relative abundance of species when the amplification efficiency is comparable for the different species. Comparison with traditional sampling methods highlights the capacity of eDNA to roughly describe the structure of a vertebrate community (Di Muri et al., 2020; Pont et al., 2018; Sard et al., 2019). Many technical factors can affect the capacity of metabarcoding to deliver “relative” quantitative results (Lamb et al., 2019), but the choice of primers, template competition and the characteristics of the mixture of species are among the most important (Piñol, Senar, & Symondson, 2019; Taylor M. Wilcox et al., 2020). Some discrepancies are related to the bias of conventional sampling methods, especially in large water bodies (Boivin-Delisle et al., 2021; Pont et al., 2018).
Several technical options have been tested to circumvent the limitation of metabarcoding to deliver absolute quantitative multiple taxa abundance. Some authors have proposed combining eDNA and animal counts (Chambert, Pilliod, Goldberg, Doi, & Takahara, 2018). Multiplex real-time PCR enables the simultaneous detection of several fishes (Jo, Fukuoka, Uchida, Ushimaru, & Minamoto, 2020). HT-qPCR systems have been tested on fish species and validated by comparison with qPCR (Wilcox et al., 2020). Simultaneous quantification of the eDNA from fish species with qSeq gives results strongly correlated with those obtained with microfluidic ddPCR (Hoshino, Nakao, Doi, & Minamoto, 2021). Another possibility (MiqSeq) is the enrichment of the sample with known quantities of DNA fragments from fish species absent from the water sample to estimate the copy number from the number of reads of local species obtained by metabarcoding (Ushio et al., 2018; Hoshino et al., 2021). To date, however, these experiments have only quantified a small number of species simultaneously and have not been tested on species-rich communities.
In this study, we propose a more direct method for inferring the absolute abundance of fish species from multiple sampling locations by combining eDNA metabarcoding with qPCR analysis, which assesses the total abundance of eDNA amplified by the universal marker used for metabarcoding. Fish-specific eDNA concentrations are then calculated from the ratios of fish species-specific read counts over the total read count of a sample (metabarcoding) multiplied by the total eDNA concentration estimated with qPCR (van Bleijswijk et al., 2020).
The effectiveness of this procedure was tested in a fish eDNA metabarcoding survey implemented along the Danube River from source to mouth (2850 km) and its major tributaries (Fig. 1). Water samples were collected from shore to shore to provide integrative sampling of the river cross section. Among the 47 sites sampled, 18 were also investigated with a conventional sampling method (traditional electrofishing, TEF) to estimate fish species abundance expressed in density or biomass per ha (Supplementary Tab. 1). We performed an eDNA metabarcoding workflow previously described (Pont et al., 2018) using the 12S mitochondrial primer for fish ”teleo” (Valentini et al., 2016). The total abundance of eDNA amplified with ”teleo” (teleo-DNA) was estimated by qPCR analysis. Our main objectives were (1) to verify the efficiency of our eDNA sampling strategy to correctly describe the fish communities and the ecological significance of longitudinal taxa profiles, (2) to evaluate the strength of the correlation between the estimated number of absolute total and specific eDNA copies per litre with the fish abundance obtained by using TEF, and (3) to model the influence of the total number of eDNA copies per sample on the taxa richness.