Geographic coordinates, elevation, lake water depth, pH and electrical conductivity were measured with appropiate devices during the different field surveys. Geographic coordinates and elevation were measured with Garmin etrex devices. Lake water depth was measured with a hand echolot (Hondex PS-7) and pH and electrical conductivity were measured with a WTW multi-340i device. Annual mean temperature, mean temperature in July and January and mean annual precipitation were downloaded from WorldClim 2 (www.worldclim.org) and are based on the average climate data for the years 1970–2000 at a spatial resolution of 30 seconds (ca. 1 km²). The site-specific climate data was interpolated to the location area by using the R packages raster (Hijmans 2020). Vegetation around the lakes was extracted from vegetation maps from Russia, published by the Institute of Geography in 1990 and China (Zhang 2007). A digital version of the Russian map is available under: https://daac.ornl.gov/cgi-bin/dsviewer.pl?ds_id=700. We extracted dominant plant taxa around the lakes, considering that most of the lakes have small basins or less-developed hydrography networks. Vegetation information was then extracted by using the buffering function in R packages raster (Hijmans 2020). The site specific ring-buffer was measured between the sampling point to the furthest lake shoreline. For some sampling points from the relatively small lakes (radius < 50 m), this buffer was set to 100 m. Dominant plant taxa were extracted and categorised in: ‘dominant_Taxon1’: indicates the most frequent taxon, ‘Taxon2-10’: Taxa listed in descending order by their area of distribution in modern vegetation. If the extracted vegetation were identified as being too diverse, no dominant taxon (‘dominant_Taxon1’) could be identified. All environmental data are summarised in Suppl. material 1

DNA extraction of surface sediment samples was carried out in the molecular genetic laboratories for environmental genetics at Alfred Wegener Institute Helmholtz Centre for Polar and Marine Research. The genetic workflow using modern surface sediment samples includes DNA extraction, amplification, purification & pooling, DNA sequencing and bioinformatic analyses. For sediment DNA extractions, about 3–5 g of wet sediment and about 8-10 g of wet sediment for samples of 13-TY, were utilised. All DNA extractions were carried out by using the DNeasy PowerMax Soil Kit (Qiagen, Germany) and PowerMax Soil DNA Isolation kit (MoBio Laboratories, Inc., USA) following the manufacturer’s instructions with the following modifications: To each sediment sample in the Power Bead solution 2mg/ml proteinase K, 0.5ml dithiothreitol (DTT) and 1.2 ml C1 buffer were added, then vortexed for 10 min and incubated overnight in a rotating shaker at 56 °C. The final elution step was carried out with 1.6 ml C6 buffer. Each DNA extraction batch contained a maximum of nine samples and one extraction blank. Precautions were taken for all the experimental steps to avoid potential contaminations (Champlot et al. 2010Epp et al. 2019). We ran separate PCR set-ups for each of the extraction batches, including a PCR negative control (NTC). We produced at least two PCR replicates of each sample, which were set up on different days. For a few samples in the AGAK-5 run, we produced up to six PCR replicates. PCR was performed using the trnL g and h universal plant primers for the short and variable P6 loop region of the chloroplast trnL (UAA) intron (Taberlet et al. 2007). For internal barcoding of the samples and PCR replicates, we used modified g and h primers with a unique 8 bp barcode and NNN was added. After evaluation of PCR results, PCR products were purified with MinElute Kit (Qiagen, Germany), including PCR NTCs. The DNA concentration of the purified products was measured with a Qubit 2.0 Fluorometer (Thermo Fisher Scientific, Germany) and finally pooled equimolarly. In total, four pools were generated and resulted in four different sequencing runs. The pool, including samples from 13-TY, was prepared differently from the other PCR pools, because here we performed PCR replicates with the same barcode combination and pooled replicates before the final equimolar pooling of all PCR products. The four pools were sequenced using the external sequencing service (Fasteris SA, Switzerland). Four independent sequencing runs were performed (Table 1). Samples presented in this study derive from four different sequencing runs. The data from the individual runs were later merged into a single dataset for the final interpretation.

We analysed in total 262 lake samples, resulting in 688 PCRs, which divide into 553 sample PCRs and 135 PCRs of extraction blanks and NTCs. The analysis of the resulting sequence data and taxonomic assignments was done by using the OBITools package (Boyer et al. 2015). Firstly, we used illuminapairedend to align the raw forward and reverse reads and kept only joined reads by using obigrep with the option -p 'mode!="joined"'. Then, we used demultiplexed the joined sequences by applying ngsfilter, which sorts the samples according to the given unique combination of internal barcodes and the primer sequences. After demultiplexing 674 PCR samples were retrieved, while from 14 PCR samples no reads were obtained. Subsequently, we used obigrep to remove sequences shorter than 10 bp and obiuniq to merge identical reads (coverage of 100% required) by keeping the total count number for each sample in which the read was detected. Further, we used obiclean to remove probable PCR and sequencing errors. For taxonomic assignment, we used ecotag with two different databases. One database comprises a sequence reference database generated from the EMBL standard sequences (http://www.ebi.ac.uk/ena) using the release 138. To apply the EMBL database with ecotag, a transformation of the EMBL into an ecoPCR database was required. Therefore, we first ran a simulated in silico PCR (Ficetola et al. 2010) with the g/h primers on the EMBL Standard Nucleotide Database (release 138) allowing five mismatches between the primers and target sequences. The resulting ecoPCR output was filtered by obigrep to ensure assignation to the taxonomic levels: species, genus and family. Then, obiuniq was used to de-replicate redundant sequences and obiconvert was applied to transform the ecoPCR output into an ecoPCR database, which is finally used by ecotag. The second database is a sequence reference database for Arctic and Boreal vascular plants (arctborbryo database), containing 1664 vascular plant taxa and 486 bryophyte species (Soininen et al. 2015, Sønstebø et al. 2010, Willerslev et al. 2014) already adjusted to the ecoPCR database format. For further analyses, we kept only those sequence reads that match 100% to the one or the other reference database. The total read sequence for each sample and its replicate(s) are plotted in Fig. 2. The full list of sequence types abundances is given in Suppl. material 3.

Figure 2.  

Compilation of total read counts for each PCR sample. As PCR replicates of 13-TY samples were not sequenced separately, but pooled together, only one total read count per sample is shown.

After merging raw paired-end reads and demultiplexing according to the internal barcode, two sample PCRs from lakes 16-KP-02-L08 and 16-KP-04-L22, as well as seven BLANKs and five NTCs yielded no read counts and were discarded from the dataset. After taxonomic assignment of the remaining 674 PCRs, we identified a total of 15,754,779 reads which had a 100% match to either the EMBL or arctborbryo database. For 99.6% of the reads (15,692,844 read counts), we identified 621 unique plant sequences types, while 0.39% (61,835 read counts) of the reads were assigned to non-plant taxa, including bacteria, algae and higher eukaryotic taxa (in total 72 unique sequence types). Further, we identified 340,346 reads (2.1% of the total dataset) in extraction blanks and NTCs, whereof 38.7% (23,975 reads) of reads in the BLANKs and NTCs were of non-plant origin. Amongst the samples, excluding BLANKs and NTCs, we found large differences in sample read counts which range between 1 and 718,279.