Comprehensive Primer Sets and Cost Ecient Multiplex PCR-based eDNA Sequencing for Community Dynamics of Cyanobacteria, Eukaryotic Phytoplankton and Zooplankton in Lake

simple and cost ecient multiplex PCR-based eDNA sequencing with effective primer sets would contribute to aquatic phytoplankton and zooplankton monitoring. The diversity of cyanobacteria, eukaryotic phytoplankton and zooplankton in Lake Tai varies signicantly from spring to summer, but shows no much geographic variation. The temperature is shown as the remarkable environmental factor affecting the diversity.


Introduction
It is a big challenge for biologists to quantify environmental biodiversity just relying on morphological and behavioral characters since the traditional identi cation is often biased, time consuming, and dependent on The taxonomic difference of all gene barcodes was compared at each of species, genus and family level respectively for each phytoplankton phylum and zooplankton ( Fig. 2-3, Figure S1-S2). The 28S marker produced the most different assignments from other markers at each of species, genus and family level for Bacillariophyta, following which the 18S-v4-5 produced the second most different species assignments from other markers (Fig. 2). For Chlorophyta, the 18S-v4-5 marker produced the most different species assignments while the 28S produced the most different genus assignments and the 16S-new produced the most different family assignments (Fig. 2). The 16S-new also produced the most different assignments at each of species, genus and family level for cyanobacteria (Fig. 3). For zooplankton, the most different assignments among all markers were from 18S-v4-5 at each of species, genus and family level (Fig. 3). For Chrysophyta and Cryptophyta, the 18S-v4-5 marker produced the most different species assignments ( Figure S1). The 28S and 16S-new markers produced the most different genus assignments and most different family assignments respectively for Cryptophyta ( Figure S1), and produced the most different species assignments for Euglenida and Pyrrophyta respectively ( Figure S2). The species assignments of all gene barcodes were also compared with the microscope observations for each phytoplankton phylum ( Figure S3). For each phylum, while some markers showed the same species assignments with the microscope observations some markers showed unique species assignments from microscope observations. For Bacillariophyta and Cyanobacteria, the microscope observations showed some unique species assignments from all gene barcodes.
Finally, all gene loci showed different assignment rate at species, genus, and family level for each phytoplankton phylum (Fig. 4). The 28S and rbcL-Cry markers showed higher species assignment rate for Chlorophyta, Bacillariophyta, Cryptophyta, Pyrrophyta, Heterokontophyta, Ochrophyta, and Zooplankton. The 23S marker showed the highest species assignment rate for Cyanobacteria. All gene barcodes showed generally higher genus and family assignment rate than species assignment rate. Community diversity patterns Page 6/29 The community diversity analysis was performed for April and August respectively. After ltering reads with low identify score the "clean reads" for samples of April and August were 4,348,278 and 4,816,816 respectively ( Fig. 5A), which were assigned to cyanobacteria, eukaryotic phytoplankton, zooplankton, fungi, few bacteria and some unassigned taxa. The proportion of various taxonomic assignments was generally consistent between April and August, where the cyanobacteria made up a large proportion followed by zooplankton, Chlorophyta, Bacillariophyta and Cryptophyta.
The top 30 species and top 30 genera from phytoplankton are shown in Fig. 5B and 5C. While the top 5 microalgae species in April were Microcystis aeruginosa (Cyanobacteria), Synechococcus rubescens (Cyanobacteria), Cyclotella choctawhatcheeana (Bacillariophyta), Teleaulax acuta (Cryptophyta), and Cryptomonas curvata (Cryptophyta) the top 5 microalgae species in August were Microcystis aeruginosa (Cyanobacteria), Cyanobium gracile (Cyanobacteria), Synechococcus rubescens (Cyanobacteria), Aulacoseira granulate (Bacillariophyta), and Cyclotella choctawhatcheeana (Bacillariophyta). It was indicated that there was a difference for the top 30 species between April and August. In April, the relative abundance of the top species in West and North was higher than that in other regions. But in August, the relative abundance of some top species was also higher in Center and East (Fig. 5B). In genus level, there was also a difference for the top genera between April and August (Fig. 5C). In April, the relative abundance of the top genera in North and West was higher than that in other regions. In August, each region of the lake showed some genera which had higher relative abundance. The top 30 phytoplankton families for the two seasons are shown in Figure S4A, and the top 30 zooplankton species for the two seasons are shown in Figure S4B. The detailed taxa identi ed and their relative abundance at each of species, genus and family level for phytoplankton from the two seasons are listed in Table S1, and the detailed taxonomic identi cation and abundance for zooplankton is listed in Table S2.

Community seasonal succession and molecular ecological network
The taxonomic assignments of all gene barcodes were combined for non-metric multidimensional scaling (NMDS) analysis, Canonical Correspondence Analysis (CCA) analysis and molecular ecological network construction as below.
The NMDS analysis was carried out for each phytoplankton phylum, zooplankton, and fungus independently. For both mock and non-mock communities, it was demonstrated that the samples in April and August were divided clearly for each of Cyanobacteria, Bacillariophyta, Chlorophyta, Chrysophyta, Cryptophyta, Euglenida, Fungi and Zooplankton (Fig. 6). The samples for Pyrrophyta from two seasons were not clearly separated despite some diversity dissimilarities. The reason was possibly that the read abundance was insu cient for clustering. However, there was not much diversity dissimilarities among samples of different regions of the whole lake in one season. For all of Cyanobacteria, Bacillariophyta, Chlorophyta, Chrysophyta, Cryptophyta, Euglenida, Fungi and Zooplankton, the samples from one season generally grouped together without obvious division.
The CCA analysis was further conducted for each phytoplankton phylum, zooplankton, and fungus respectively, to explain the impact of environmental factors to their community composition. Samples that were ampli ed by different gene barcodes were analyzed separately to avoid sample bias, which means that the mock community was analyzed independently (Fig. 7, Figure S5). For all of phytoplankton, zooplankton, and fungi, it was shown that the temperature (WD) was remarkable in affecting community diversity variation among the whole lake from April to August with the longest vector for both mock and non-mock communities, where the samples in the two seasons were separately clearly (Fig. 7, Figure S5). However, the samples in one season clustered together without apparent dissimilarity for all taxa analyzed. There was close correlation for temperature and the total nitrogen (TN).
The molecular ecological network of assigned taxa from the mock community was constructed for April and August separately (Fig. 8). After ltering taxa with less abundance by standard parameters, the various taxa from Cyanobacteria, Chlorophyta, Chrysophyta, Bacillariophyta, Cryptophyta, zooplankton, fungi, few bacteria, and some un-assignments were included for molecular network construction. It was indicated that the most interactions among all taxa included were positive for samples in both April and August.
Particularly for samples in August, there was only one negative correlation between two unassigned taxa.
For samples in April, interactions among some taxa in Chlorophyta was negative.

Comparison of singleplex and multiplex PCR-based eDNA sequencing
The assignment consistency of singleplex and multiplex PCR was veri ed by observing the rate of their shared assignments to their total assignments. It was shown that there was assignment dissimilarity at each of species, genus and family level between singleplex PCR and multiplex PCR for most of the gene barcodes from 6 samples in the mock community (c1, c2, c3, c4, c5, c6) ( Figure S6). We also found that some gene barcodes in the multiplex PCR systems ampli ed more taxa than in the singleplex PCR system, but some gene barcodes in the singleplex PCR system also ampli ed different taxa from the multiplex PCR system. Cross ampli cation existed between singleplex and multiplex PCR reactions.

Discussion
The aquatic environment exhibits enormous microbial diversity which often interacts with each other to affect the ecosystem. The enormous microbial diversity brings major challenges to model microbial systems and to explain patterns of community variation across space and time, especially for lakes with serious environmental problems like bloom. Yet, our understanding about the aquatic microorganism diversity remains in its infancy, especially for the eukaryotic organisms while the bacteria and fungi monitoring have been already triggered by NGS. Take Lake Tai as an example, here we developed simple and cost-e cient multiplex PCR-based eDNA sequencing strategy and multiple new gene barcodes for revealing diversity patterns and seasonal community dynamics of microbial community diversity with lake algae bloom.
Gene barcodes for phytoplankton and zooplankton eDNA sequencing For both traditional DNA barcoding and eDNA metabarcoding, conserved primers of gene barcodes is useful for detecting a wide range of targets with broad sensitivity, such as the primers of 16S, ITS and 18S which are well-tested for obtaining a relatively wide range of microbial identi cation 17,71−72 . But even for the conserved primers, they do not always have equal a nity for all possible DNA sequences since different species often vary in these primer-ampli ed regions, which could consequently induce bias during PCR ampli cation 71 . So multiple gene barcodes and speci c primers would greatly contribute to microbe diversity monitoring.
In this study, we focused on the diversity analysis of prokaryotic and eukaryotic microalgae and zooplankton, especially for bloom microalgae. Firstly, we aimed to compare the identi cation e ciency of multiple gene barcodes at each of species, genus, family and phylum level for phytoplankton and zooplankton, especially for ve gene loci newly designed. It was clear that the 9 gene barcodes produced different assignments from each other, especially for the new markers. The 16S-new could be as effective candidate gene for identifying cyanobacteria since it assigned many different taxa from 23S which specially identi es cyanobacteria. But the 16S-new primers also obtained sequences longer than 600 bp. We will focus on designing primers for amplifying shorter 16S-new fragments for compatibility with NGS sequence platforms. The 28S was effective in identifying phytoplankton and zooplankton, especially for Bacillariophyta, which could be universal marker for identifying aquatic eukaryotic organisms. For three regions of 18S, the new 18S-v4-5 assigned relatively even proportion for various groups, and produced different taxonomic assignments from 18S-v1-3 and 18S-v9 at each of species, genus and family level. We had tried to use available 18S-v4 primers to amplify the samples 57 , but they failed. Among the three regions of 18S, the 18S-v4-5 region is more variable for identi cation. In future studies, we will continue to optimize 18S-v4-5 primers for obtaining fragments shorter than 500 bp. The 18S-v9 region has been usually used for identifying eukaryotic organisms by NGS 32 . In this study, the 18S-v9 was proved useful for identi cation of eukaryotic phytoplankton and zooplankton. Compared with 18S-v4-5 and 18S-v9, the 18S-v1-3 was more effective in distinguishing zooplankton. We also designed speci c rbcL primers which speci cally ampli ed Chlorophyta and Cryptophyta successfully and produced high-resolution in species identi cation. In conclusion, we prove that it is important to combine multiple gene barcodes to get comprehensive taxonomic diversity for microalgae and zooplankton.
Finally, the microscope species identi cation was compared with the taxonomic assignments of each gene marker for each phytoplankton phylum. The microscope observation detected some species of Bacillariophyta and Cyanobacteria which were not identi ed by multiple gene barcodes, which is because there is no any sequence of the gene barcodes deposited in the public databases for these species as references. We combined ecoPCR and blastn to get relatively complete reference database. But there were still some sequences which could not be assigned to any taxa due to the incomprehensive reference sequences in public database. So it would be important to supplement as many as marker sequences into the public database to form a comprehensive reference database 72 . On the other hand, the multiple gene barcodes assigned many species that the microscope observations did not discover. These species undiscovered by microscope are possibly cryptic species or species that are very di cult to be distinguished by tiny morphological characters 69-70 . Community diversity patterns and succession for algae bloom The cyanobacterial blooms in Lake Tai have led to serious environmental and societal problems, with longterm negative impacts on water quality, sheries, aesthetics, tourism and other economic activities 55,58,60 .
Most studies about Lake Tai focused on the impact of environmental factors to cyanobacterial blooms 64,66 where the microscope observation was often used for phytoplankton identi cation. The lack of conserved genes for identifying phytoplankton makes it signi cant to employ multiple gene loci for phytoplankton barcoding. Zooplankton identi cation by barcoding has already been conducted for Lake Tai with single COI gene 65 . But it would be important to comprehensively understand zooplankton diversity among the whole lake from different seasons and its interaction with phytoplankton.
Based on the multiple gene barcodes, we aim to reveal the diversity patterns, spatiotemporal dynamics and molecular network of phytoplankton and zooplankton communities in Lake Tai, in combination with the impact of environment factors. Firstly, the diversity analysis at each of species, genus, family and phylum level was performed. The proportion of each group assigned between April and August was generally consistent, but the total reads of samples in August was higher than that in April for the same sample size, which indicates that the species abundance in summer is larger than spring. The top phytoplankton taxa at each of species, genus and family level between April and August were also different between April and August. These diversity variations suggest that the growth of phytoplankton changes with the time.
However, there was no apparent difference about the diversity composition from spring to summer. The heatmap analysis showed that the diversity of various top taxa was slightly different among the ve regions of the lake in one season.
The NMDS results indicated that the samples between April and August were separated clearly for each phytoplankton phylum, zooplankton and fungi, but the samples within one season were not separated clearly. These suggest that the diversity composition of phytoplankton and zooplankton in Lake Tai varies from spring to summer. Then what could cause the diversity variance of phytoplankton, zooplankton and fungi from different seasons? The CCA analysis was conducted for each phytoplankton phylum, zooplankton and fungi, which indicated that the samples from April and August were also separated clearly for all of phytoplankton, zooplankton and fungi. The temperature was shown as the remarkable environmental factor affecting the diversity variance among temperature, TP, TN, NH4 + -N, and NO3 − -N. So the CCA analysis could give an explanation that the diversity of phytoplankton and zooplankton in April is signi cantly different from August since the temperature in summer is much higher than spring. The satellite data of 11 years from the lake showed that high temperatures and nutrient concentrations in spring promoted cyanobacterial growth 56 . Although the detailed bloom mechanism is not determined it has been pointed that the temperature stimulates cyanobacteria bloom in several ways like nitrogenase activity and growth rates 56,73−80 . Additionally, from our results, we point that except cyanobacteria the temperature could also affect the diversity variance of eukaryotic phytoplankton and zooplankton. Finally, the molecular association network showed that the interactions among most taxa of phytoplankton, zooplankton and fungi were positive for samples in both spring and August. This is possibly because the taxa among the different taxonomic groups have positive correlations in the aquatic food chain 55 . While microalgae grow quickly their grazer can also have high biomass. In future studies, we will associate the bacteria diversity and perform biodiversity monitor for long time of years with more environmental factors involved to better understand the bloom.
Strategy of multiplex PCR-based eDNA sequencing for microbial diversity While multiple gene barcodes are used in high-throughput amplicon sequencing the cost of amplicon library preparation is also multifold in target PCR, DNA cleanup, PCR enrichments, DNA quantity and index and adaptor ligation, especially for large samples. Except the cost, the laboratory work is also very timeconsuming for multiple gene loci. Multiplex PCR is similar to the singleplex PCR except that each sample is designed to amplify and detect multiple target sequences rather than only a single target, which makes it possible to prepare the multiple amplicon NGS libraries once while the multiple targets are ampli ed together. However, multiplex methods must deal with interactions between multiple sets of primers that may have different annealing temperatures and may cross-react and generate primer dimers. Henrik et al. Here we developed multiplex PCR-based amplicon Illumina sequencing in environmental microbial diversity.
Firstly, we selected and tested abundant primers of various gene loci to ensure the PCR e ciency of single locus. Then we optimized the annealing temperatures under which the multiple gene loci could be ampli ed successfully in a reaction mixture, without cross-reaction targets and apparent primer dimers. Finally, three multiplex PCR systems were constructed for the 9 gene loci since it would be impossible to combine all the gene loci in one reaction mixture because some of them had different annealing temperatures. So based on the two-step library preparation, we used a total of three locus-speci c PCRs, three cleanup reactions and one indexing PCR for 9 gene loci of one specimen, saving at least 6 cleanup reactions and 6 DNA quantities for each specimen, at a total cost of < 5 USD due to reagent difference. By Illumina Miseq sequencing, about 100000 raw reads were generated for each specimen in this study, costing about 6-8 USD to recover all loci.
The sequence cost depends on the choice of sequence yield. Thus, our multiplex PCR-based amplicon NGS sequencing make it possible to generate nine locus dataset for a total cost of < 4 USD with sequence yield of 50000 raw reads per specimen.
However, the taxonomic difference existed between singleplex and multiplex PCR, which means that their taxonomic assignments were not completely consistent. We found that both multiplex and singleplex PCR reactions generated some unique assignments. There was cross-identi cation between singleplex and multiplex PCR strategies. PCR bias affecting amplicon accuracy is inevitable, like DNA Polymerase, PCR cycle number, template concentration and annealing temperature, even for single amplicon PCR 1 . Even the singleplex and multiplex PCR had almost same PCR conditions their locus primers were signi cantly different and thus could affect the PCR reaction. Whatever, the multiplex PCR provides cost-e cient, quick and accurate amplicon NGS sequencing despite some bias. We will further focus on optimizing the multiplex PCR condition for assignment accuracy.

Sampling, microscope identi cation and DNA extraction
The Lake Tai was divided into 5 regions based on the reported plankton distribution experience ( Figure S7 (Table 1). Due to PCR failure of available 18S-v4 primers, new 18S-v4 primers were designed. New primers of 28S were also designed. We also designed primers for rbcL which could be a speci c marker to obtain Chlorophyta and Cryptophyta. After downloading related reference sequences from EMBL the program DegePrime was used to design potential primers 81 . Multiplex PCR and Library preparation Firstly, we ampli ed each locus individually using a Qiagen Hot-Start PCR kit to con rm that they will produce only one single clear fragment, as visualized after agarose gel electrophoresis. PCR reactions were carried out in a total volume of 10 µL, using 5µL Master Mix, 1µL DNA template (2-10 ng/µL), and 0.1 µM of forward and reverse primer. Then multiple pairs of locus-speci c primers which ampli ed single clear fragment were mixed together for gradient PCR (annealing temperatures from 45 °C to 68 °C) to select optimal multiplex PCR conditions under which multiple gene loci could be ampli ed together with same annealing temperatures and without cross-reactions. After optimization, four gene targets (18S-v1-3, 18S-v9, 23S and 16S) could be ampli ed in one PCR reaction (A) with annealing temperature 63 °C, two gene targets (rbcL-Chlo and rbcL-Cry) could be ampli ed in one PCR reaction (B) with annealing temperature 55 °C, and two gene targets (28S and 18S-v4) could be ampli ed in one PCR reaction (C) with annealing temperature 55 °C (Table 1). Multiplex PCR reactions were carried out in a total volume of 25µL with Qiagen Multiple PCR Plus kit, using 12.5µL Master Mix, 1µL DNA template (2-10 ng/µL), 0.1 µM of forward and reverse primer, and Q-Solution. PCR conditions for the three multiplex PCR systems were: 95 °C for 5 min, 30 cycles of 95 °C for 30 s, primer-speci c annealing temperatures for 1 min 30 s, 72 °C for 1 min 30 s, with a nal extension of 72 °C for 1 min. Based on the optimized multiplex PCR conditions, two-step NGS library preparation was performed subsequently. Primers were synthesized with the locus-speci c sequence on the 3' end and a 5' tail containing sequence matching the TruSeq sequencing primer binding site (The forward primer was synthesized with a 5' tail of ACACTCTTTCCCTACACGACGCTCTTCCGATCT and the reverse primer with a 5' tail of GTGACTGGAGTTCAGACGTGTGCTCTTCCGATCT).
The rst PCR was conducted with these tailed primers using annealing temperatures 55 °C for PCR system A and B, and 49 °C for PCR system C. Triple PCR products for each sample were pooled, which was then cleaned by Solid Phase Reversible Immobilization (SPRI) Beads. After puri cation, the cleaned PCR products were quanti ed. Then equal amounts of PCR products from the three multiplex PCR systems for each sample were pooled for each sample. Indexing PCRs were performed for each pooled amplicon with indexing primers annealing to 5′-tails of the locus-speci c PCR primers. The index PCR conditions was: previously identi ed by microscope. The ve new gene barcodes designed in this study were ampli ed in the mock community for comparing identi cation e ciency of all gene loci and microscopic observations. For comparing the assignment consistency between singleplex and multiplex PCR, we selected 6 samples to perform the single PCR for each locus per specimen, where the PCR conditions were the same as multiplex PCR.

Bioinformatics analysis
The ecoPCR 82 was rstly run to build the original database for each gene marker on the EMBL standard nucleotide database from both std and wgs 83 . For each marker, 3 mismatches was set for silico PCR. As the CRUX module in Emily (2019) 30 reported, blastn 84 was then used to query the seed databases against the NCBI non-redundant nucleotide database to increase the breadth of reference sequences and capture sequences without barcode primers in EMBL 85 . Taxonomy les were retrieved using Entrez-qiime 86 . We also added our previous published phytoplankton sequences from traditional DNA barcoding into the databases for 16S and rbcL. Finally, if there were still reads which could not be assigned to any taxa we blasted them against the whole NCBI nucleotide database again to get the unique perfect-matched species assignment.
Sequence analysis for all gene barcodes was performed using QIIME2 87 . Raw sequence fastq les were rstly demultiplexed to separate the samples based on the indexes. PCR primers were removed by cutadapt. Dada2 was applied for denoising, dereplicate-sequence, ltering chimera and merging the paired reads. For 18S-v1-3, 18S-v4-5 and 16S-new, paired reads merged with good quality score were used or only forward reads were used if the reads could not be perfectly jointed. After quality lter, taxonomic classi cation was performed by feature-classi er module. For reads that were identi ed as unsigned we blasted them against the whole NCBI (nt/nr) database to get the unique perfect-matched species assignment with the highest identity score 1. Finally, the taxonomic assignments and reads abundance for all gene loci were obtained for biodiversity analysis.
The mock community was used to compare the identi cation e ciency of all gene loci. The proportion of different taxa for each marker was counted at the phylum level. Then at each of species, genus and family level, the taxonomic dissimilarity of all markers for phytoplankton and zooplankton was compared by Venn package in R, where the comparison was performed for each phytoplankton phylum, including Cyanobacteria, Chlorophyta, Bacillariophyta, Cryptophyta, Pyrrhophyta, Euglenophyta and Chrysophyta. The assignment rate of each gene marker at each of species, genus and family level was also performed respectively for each phytoplankton phylum, which meant the proportion of reads assigned to each taxonomic level to total reads from assignments of each marker for each phytoplankton phylum.
Taxonomic assignments from all gene loci were combined for biodiversity analysis. The community diversity analysis of two seasons was conducted respectively from the total ASV (amplicon sequence variant) of all samples including the mock community, which was performed by heatmap implemented in R package. For all samples, pairwise community dissimilarity was calculated using Bray-Curtis as implemented in the vegan package 88 . The non-metric multidimensional scaling (NMDS) plots based on Bray-Curtis distance was generated to reveal the diversity dynamics of samples among the whole lake in April and August respectively by vegan 88 , which was performed for each phytoplankton phylum and all of zooplankton. The Canonical Correspondence Analysis (CCA) analysis was performed to evaluate the correlations of environmental factors and diversity dynamics of samples among the whole lake in Aril and August, in the correspondence analysis by calling the "cca" function from vegan package 88 . The CCA analysis was also conducted for each phytoplankton phylum and all of zooplankton, where the samples ampli ed by different gene loci were analyzed independently to avoid sample bias. Permutation tests were performed to evaluate the signi cance of overall models. The vegan package was employed in R 3.6.1. For understanding the interactions among various taxa, the ecological association networks were constructed by molecular ecological networks (MENs) models in Deng (2012) 89 . The cytoscape was used to visualize the networks (https://cytoscape.org/). For obtaining comprehensive associations, the samples from the mock community were used for the molecular network construction.

Conclusions
Comprehensive and accurate identi cation of aquatic microbial diversity is important to understand marine and freshwater ecology, especially for algae blooms. The comprehensive primer sets of 9 gene markers among 18S, 28S, 23S, 16S and rbcL, including our newly designed ones, assign cyanobacteria, eukaryotic phytoplankton and zooplankton differently at each of species, genus, family and phylum level, which suggests that it is critical to employ multiple effective gene markers for aquatic accurate microbial diversity detection. Our newly developed simple and cost e cient multiplex PCR-based eDNA sequencing with effective primer sets would contribute to lake phytoplankton and zooplankton monitoring. The cyanobacteria is dominant in the phytoplankton, followed by Bacillariophyta and Chlorophyta in both spring and summer in Lake Tai. The phytoplankton and zooplankton diversity varies signi cantly from spring to summer, but shows no much geographic difference among the lake. Most interactions among phytoplankton and zooplankton are positive in the molecular ecological network. The temperature is shown as the remarkable factor affecting the community dynamics. These results suggest that the temperature possibly plays key role in algae bloom. In future studies, we will associate the bacteria diversity and perform microbial biodiversity monitor for long time of years to better understand the lake microbial diversity and algae bloom.

Declarations Ethics approval and consent to participate
This study is not involved in human participate, human data and human tissue.

Consent for publication
This study does not contain any individual person's data in any form.
Availability of data and materials Figure 1 The proportion of taxa assigned from each gene marker.
Page 23/29 Figure 2 The difference of taxonomic assignments among all gene barcodes at each of species, genus and family level for Bacillariophyta and Chlorophyta. Red: species level; Cyan: genus level; Lilac color: family level.   Canonical correspondence analysis (CCA) ordination diagram for correlations of diversity dynamics and environmental factors for each phytoplankton phylum, zooplankton, and fungi. Samples analyzed were ampli ed by gene barcodes from PCR system A (non-mock community). The length of the vector is proportional to its importance, and the angle between two vectors re ects the degree of correlation between variables.

Figure 8
The molecular network among various taxonomic assignments for April and August respectively. The connection lines in gray and green indicated the positive and negative correlations respectively.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.