Get Complete Project Material File(s) Now! »

Material and methods

Study site and coring

Lake Savine is located in the Haute-Maurienne massif of the French-Italian Alps (2447m a.s.l., Figure 3-1). The lake’s catchment has a 3.5 km2 surface, rising up to 3310 m. a.s.l. and is mainly (90%) made up by crystalline rock with Permian gneiss and micaschist. The lake receives detrital inputs mainly from the southeastern part of the catchment from mid-June to mid-November, while the catchment is covered with snow and the lake is frozen the rest of the year. Geological and coring details are described in (Sabatier et al., 2017). Today, the lake’s catchment area is covered by high altitude alpine meadows and siliceous screes. Nowadays, the meadows are occasionally grazed by cattle during summer.
In March 2014, a 775-cm-long sediment core was extracted from the deepest part of Lake Savine (7m depth, N45°10.500, E6°54.821). Sabatier et al., (2017) presented a high-resolution sedimentological and geochemical analysis of the 775-cm-long sediment core sampled in Lake Savine, which led to the identification of 200 flood event layers and 20 underwater mass movements possibly related to earthquakes for the last 6500 years. The chronology of this sediment core results from 12 14C ages on terrestrial organic macro-remains and short-lived radionuclides (see Figure 3-S1 for the age model), for further detail see Sabatier et al., (2017).

Extraction, amplification and sequencing of extracellular DNA

From the sediment core, we sampled 72 slices (2-cm thickness) covering the whole core. Details of the sampling scheme (depths, thicknesses, weights) can be found in Table 3-S1. In order to avoid potential contamination, we removed the edges of slices with decontaminated, disposable scalpels. DNA extraction targeted extracellular DNA (Pansu, Giguet-Covex, et al., 2015). For each sediment slice, we mixed about 15 g of sediment with 15 ml of saturated phosphate buffer (Na2HPO4; 0.12 M; pH ≈ 8) during 15 min. 15 ml of the mixture was then centrifuged (10min at 10000 g). 12 ml of the resulting supernatant were transferred to Amicon ® Ultra-15 10K Centrifugal Filter Devices (Millipore) and centrifuged (20 min at 4000 g) for ultrafiltration and concentration of DNA. 400 µl of the resulting concentrate was kept as starting material for the following extraction steps, using the NucleoSpin® Soil kit (Macherey-Nagel, Düren, Germany), skipping the cell lysis step and following manufacturer’s instructions [as described in (Taberlet, Prud’Homme, et al., 2012)].
Mammals and plants were the targets of DNA amplification. Mammal DNA was amplified with a primer pair targeting on a 60 to 84bp fragment of the mitochondrial 16S gene (P007 in Giguet-Covex et al., 2014; which corresponds to Mamm02 of Taberlet, Bonin, Zinger, & Coissac, 2018). For plant DNA, we used the primers targeting the P6 loop region of chloroplast trnL intron, a fragment whose length varies from 10 to 220 bp (g-h in Taberlet et al., 2007, which corresponds to Sper01 in Taberlet et al., 2018). To assign sequence reads to the relevant sample, 8 bp tags (with at least 5 differences between them) were added to the 5’ end of primers (Binladen et al., 2007; Valentini, Pompanon, & Taberlet, 2009). DNA amplifications were carried out in a final volume of 20 µL containing 2 µL of diluted DNA extract. The amplification mixture contained 1.2 U of AmpliTaq Gold® DNA Polymerase (Applied Biosystems), 15 mM Tris-HCl, 50 mM KCl, 2 mM of MgCl2, 0.2 mM of each dNTP, 0.2 µM of each primer and 4.8 µg of bovine serum albumin (BSA, Roche Diagnostic). To limit the amplification of human DNA when using Mamm02 primers, we also added human-specific, blocking oligonucleotide (MamP007_B_Hum1) After 10 min at 95°C, we performed 45 PCR cycles: 30 s at 95°C, 30 s at 50°C, and 1 min at 72°C with a final elongation step (7 min at 72°C) was performed. In addition to extraction controls, we also performed three PCR controls, containing PCR mix but no DNA template, and four PCR positive controls, each containing 0.18 ng of DNA extracted from a marsupial (Didelphis marsupialis, Didelphidae), absent in Europe. The positive control for plants was extracted from Mouriri sp., a tropical plant from the South America, belonging to a family (Melastomataceae) also absent in Europe. All samples and controls were amplified in 12 replicated PCRs (Ficetola et al., 2015). PCR products were then purified and mixed before sequencing. Sequencing was performed by 2 x 125 bp pair-end sequencing on Illumina HiSeq 2500 platform.

DNA sequence read filtering and taxonomic assignment

We filtered the DNA sequences adopting the protocol described in Pansu et al., (2015), using the OBITOOLS software package (Boyer et al., 2016). First, paired-end sequences were aligned and merged using the illuminapairedend program which implements a gap-free alignment algorithm. Identical sequences were merged together with their information of original samples being kept by using the obiuniq command. Merged sequences with alignment score lower than 40 were discarded. Each retained sequence was assigned to its original sample by the ngsfilter program. A sequence was retained only if it contained both primers with no more than two mismatches in each as well as the exact sample tag sequence. Using the obigrep command, we discarded all sequences shorter than 10 bp or containing ambiguous nucleotides. Sequences occurring only once in the dataset were discarded. To filter out potential amplification and sequencing errors, the obiclean program were used to create networks of sequences by linking any sequences differing by a single nucleotide substitution or indel. According to it corresponding network, each sequence fell into one of three categories: ‘heads’, which were only connected to sequences with a lower count, ‘singletons’, which were not connected to any other sequence and ‘internals’, which did not fall into either of the categories above. Sequences falling more often into ‘internal’ than ‘head’ or ‘singleton’ were then considered to be likely erroneous sequences and discarded.
Each remaining sequence formed a molecular operational taxonomic unit (MOTU). Taxonomic labels were assigned to each MOTU with the ecotag program by aligning them to a reference database. The database was generated by in silico PCR using the ecoPCR program (Ficetola et al., 2010) on the EMBL release 128 nucleotide database (nt). For plants, the taxonomic alignment was firstly performed on the global database including all taxa in the EMBL database and then on a database with only alpine species (Pansu, Giguet-Covex, et al., 2015). Inconsistences between the two resulting assigned datasets were resolved based on knowledge about the present local vegetation (Table 3-S3). We consider PCR replicates with more than 10 reads as positive detection, given that we consistently obtained much more plant than mammal reads. Only those MOTUs assigned to family or better taxonomic levels (genus, species etc.) with a perfect match (100%) and with more than 4000 reads and more than 10 positive PCR replicates over the whole dataset were retained. For mammals, we kept only the MOTUs with a match higher than 97% to a mammal genus and with more than 5 reads in a sample (Sabatier et al., 2017). MOTUs with more than three positive replicates in the control samples were removed from the dataset. MOTUs assigned to exotic taxa were obvious contaminants and also discarded (Table 3-S2). The remaining MOTUs were kept for data analyses.


Site occupancy modeling

In SedDNA studies, both false absences (i.e. failure to detect a taxon when it is present) and false presences (i.e. detection of a taxon when it is actually absent, for instance because of contamination) are possible (Ficetola et al., 2015). Correctly inferring the presence of domestic animals is crucial in order to explore the relation between human activities and vegetation change. We adopted a site occupancy modeling approach to estimate the probability of a sediment sample being occupied, given the number of detections in each sample, using the Misclassification Model (Royle & Link, 2006):
where ψ is the occupancy probability for all samples; zi the state of occupancy in sample i (zi = 1 when the site is occupied, otherwise zi = 0); yi the number of positive detections in sample i; p11, p10 the true and false detection probabilities, respectively. K denotes the number of observations and is regarded as number of PCRs in this study (Ficetola et al., 2015). The conditional probability that a detection history with detections comes from an occupied sample can be calculated by (Lahoz-Monfort et al., 2016):
The model was coded in R and feeded to JAGS version 4.2.0 (Plummer, 2003) through the R2jags package (Su & Masanao Yajima, 2015) to obtain the posterior probability distributions of ψ, p11, p10 and Pi using 3 MCMC chains of 10000 iterations for each MOTU. The Gelman-Rubin statistic for each parameter was computed to check chain convergence (Kéry, 2010).

Ecological analyses

We based our ecological analysis on a plant DNA dataset consisting of presence/absence values of the MOTUs remaining after the filtering step. The oldest samples from the core (up to ca. 8300 cal. years BP) were excluded because of lack of plant DNA. DNA detected from the remaining samples was transformed to presence/absence value and then combined for further analysis.

Table of contents :

1.1 Biodiversity and ecosystem functioning in the Anthropocene
1.2 eDNA metabarcoding for the study of past biodiversity
1.3 Workflow in eDNA metabarcoding for past biodiversity study
1.4 Ecological analysis for sedimentary DNA metabarcoding data – quantitative approaches
1.4.1 Defining the operational units of eDNA data for ecological analysis
1.4.2 Site occupancy-detection models
1.4.3 Measuring abundance
1.4.4 Quantifying species diversity
1.4.5 Partitioning samples to find hidden structure
1.4.6 Finding main trends by ordination
1.4.7 Testing the effects of environmental variables on community composition
1.4.8 Testing complex causal relationships
1.5 Reference
2.1 Introduction
2.2 Materials and Methods
2.2.1 Analysis of the literature
2.2.2 CAR-SODM
2.2.3 Parameter estimation
2.2.4 Simulations: chronosequences
2.2.5 Simulations: spatially autocorrelated dataset
2.2.6 Model performance comparison
2.2.7 Analysis of Empirical data
2.3 Results
2.3.1 Literature analysis
2.3.2 Simulations: chronosequences
2.3.3 Spatial data simulations
2.3.4 Analysis of empirical eDNA data
2.4 Discussion
2.5 Conclusions
2.6 Acknowledgements
2.7 Data accessibility
2.8 Author contributions
2.9 Reference
2.10 Supporting Information
3.1 Introduction
3.2 Material and methods
3.2.1 Study site and coring
3.2.2 Extraction, amplification and sequencing of extracellular DNA
3.2.3 DNA sequence read filtering and taxonomic assignment
3.2.4 Site occupancy modeling
3.2.5 Ecological analyses
3.3 Results
3.3.1 SedDNA dataset
3.3.2 Temporal changes of communities
3.3.3 Environmental predictors of community changes
3.3.4 Disturbance and temporal variability of communities
3.4 Discussion
3.4.1 Vegetation dynamics in high altitudes
3.4.2 Role of climate change
3.4.3 Impacts of livestock farming
3.5 Acknowledgements
3.6 Data Accessibility
3.7 Author Contributions
3.8 Conflict of interests
3.9 References
3.10 Supplementary Information
3.10.1 Supplementary tables and figures
3.10.2 Testing the impact of flood deposit content on DNA detection
4.1 Introduction
4.2 Results
4.2.1 Sedimentology and chronology
4.2.2 DNA dataset
4.2.3 Plant community variation
4.2.4 Occurrences of domestic animals
4.2.5 Erosion rate
4.2.6 Interactions among climate, livestock farming, plant community composition, and soil erosion
4.3 Discussion
4.4 Material and methods
4.4.1 Study site
4.4.3 Estimating the erosion
4.4.4 Coprophilous fungi spores
4.4.5 Statistical analyses
4.5 Acknowledgements
4.6 Data Accessibility
4.7 Author Contributions
4.8 Reference
4.9 Supporting information
4.9.1 Supplementary discussion on plant community change


Related Posts