A total of eight water samples were collected in September and November 2019, at four separate sampling sites in Qiantang River lower reaches, including Tonglu, Fuyang, Wenyan and Jiashao sites, hereafter referred to as TL, FY, WY and JS separately (Figure 1). All sampling, filtering, and other equipments were sterilized before use.

2 L of surface water was collected at a set time by boat from each site. Simultaneously, five environmental variables, such as transparency, salinity, water temperature (WT), pH and dissolved oxygen (DO), were measured in situ by using a secchi disk and a portable water quality detector (Hach, USA). For each

water sample, a 1.5 L water subsample was used to test six environmental variables immediately after transported to the laboratory within 24 hours at 4˚C, including Chemical oxygen demand (COD), Chlorophyll a (Chla), total nitrogen (TN), total phosphorus (TP), ammonium (NH₄-N) and nitrate-nitrite (NO₃-N), the recording of which were following the scheme of [23]. All environmental variables were measured in triplicate.

For each water sample, a residual 500-mL water subsample was then filtered through a 0.22-μm cellulose acetate filter paper using a peristaltic pump in the field. Then, each paper was placed inside a commercial sterile centrifuge tube and stored in a container filled with liquid nitrogen until subsequent manipulations were performed. DNA was extracted from filters using EZNA water DNA kit (Omega, USA) following the manufacturer’s protocol. The concentration and purity of DNA were determined using NanoDrop 2000c spectrophotometer (Thermo, USA), then followed by multiplex PCR using the universal primers for ITS belonging to eukaryotic mitochondrial DNA fragments, ITS-F (5'-GTGA ATCATCGARTC-3), ITS-R (5'-TCCTCCGCTTATTGAT-3') [22]. Each eDNA sample was subjected to PCRs in triplicate [17.35 μL dH₂O, 2.5 μL 10 × PCR buffer, 0.15 μL 5 U/μL Thermo scientific Taq DNA polymerase, 2 μL dNTPs mix (2.5 mM), 1 μL of each primer (10 mM), and 1 μL sample eDNA in a total volume of 25 μL]. PCR cycling parameters were as follows: 96˚C for 3 min, followed by 35 cycles of 96˚C for 30 s, 50˚C for 30 s, and 72˚C for 90 s, with a final elongation step at 72˚C continued for 7 min. Negative control was conducted simultaneously. After subjecting the PCR products to 1% polyacrylamide gels (see FigureS1), the quantified, size-selected libraries were constructed and continuously diluted to a concentration suitable for sequencing. The libraries were finally sequenced on the Illumina MiSeq 2000 platform by following the manufacturer’s protocols step by step.

Phytoplankton samples were also simultaneously collected at the four sites. For phytoplankton counts, 1.0 L of water samples were sampled each time and preserved with 1% Lugol’s iodine solution. Phytoplankton samples were concentrated to a final volume of 30 ml after sedimentation for 48 h. Thereafter, the taxa were verified and counted under 200× and 400× magnifications for at least 500 specimens [23]. The data were made to compare with that collected from ITS gene metabarcoding method.

The raw sequencing FASTQ file was transformed to a FASTA file by the Fastx toolkit V0.0.1 [24]. Clean reads were gained after trimming the low quality sequences and PCR chimeras by using Fqtrim V0.9.4 (http://ccb.jhu.edu/software/fqtrim/) and Vsearch 2.3.4 [25]. Vsearch 2.3.4 software continued to be utilized to cluster Operational taxonomic units (OTUs) at 97% cutoff of sequence similarity. Representative sequences of OUTs were assign to taxonomic groups against the NCBI Genbank by using Blast+ 2.6.0 at 80% cutoff of identity thresholds with over 80% matches and expect values less than 1e−5. The Sequencing data have been uploaded to NCBI Sequence Read Archive database, with accession numbers vary from SRR10800795 to SRR10800802.

Three α-diversity indices, including Chao1 estimators, Simpson index, and Shannon index, were calculated based on data obtained by metabarcoding monitoring. Additionally, Cluster analysis taken by group average method and Non-metric multidimensional scaling (NMDS) was employed to cluster samples in Primer 5.0 environment [26], of which species data were first transformed according to [27]. Additionally, basic data processing, drawing and statistical analyses (e.g. one-way ANOVA) were conducted using Excel 2007 and SPSS 16.0 software.

The results of environmental variables are showed in Table 1. WT showed significant differences between months. TN showed significant negative association with transparency at p < 0.05, with COD at p < 0.01, and significant positive association with TP at p < 0.01. In addition, NH4-N showed a significant positive association with TP and WT at p < 0.05. The values of TN: TP mass ratios were all higher than 7 in all sampling sites, indicating the research area was generally P-limited at the experimental period.

ITS gene metabarcoding yielded 67,469 - 129,150 raw reads, of which 55,687 - 112,832 clean reads were obtained after optimization, resulting in effective data rates varying from 64.1% to 88.8% (Table 2). The sequences clustered into a total of 5795 OTUs, varying from 706 to 1911 at an average of 1245 (Table 2). Meanwhile, the rarefaction curves of each sample all showed the observed species number flatted out as sequence increasing, indicating the amount of sequencing data at the 97% similarity threshold was sufficient to satisfy the assessment of species diversity.

In total, Phytoplankton, fungi, zooplankton and other eukaryotes were annotated on after Blast. Five classes of eukaryotic phytoplankton were annotated on,

A: the “September” sample; B: the “November” sample. Hereinafter inclusive; +: salinity measured; −: no salinity measured.

Sep.: September; Nov.: November.

in which Chlorophyceae, Trebouxiophyceae and Cryptophyceae had more reads abundance (Figure 2(a)). Genera Chlamydomonas, Micractinium, Chlorella, Crucigenia, Cryptomonas, Actinastrum, Gonium,Dictyosphaerium and Compactochlorella were the common phytoplankton, most of which belong to phylum Chlorophyta, except Cryptomonas which belongs to phylum Cryptophyta. Meanwhile, three classes of zooplankton, including Ciliatea (Protozoa), Maxillopoda (Arthropoda) and Hydrozoa (Cnidaria), were annotated on, and the first two classes were the common zooplankton assemblages (Figure 2(b)). Furthermore, twenty classes of fungi were annotated on, Dothideomycetes, Sordariomycetes, Eurotiomycetes and Tremellomycetes were the common classes (Figure 2(c)). However, the common classes of fungi in JS-A sample were others, such as Malasseziomycetes, Agaricostilbomycetes, Ustilaginomycetes and Wallemiomycetes.

Overall, the total reads of ITS annotated on eukaryotic microplankton in September was higher than that in November. The dominant assemblage in TL, FY and WY sites in September was phytoplankton, contributing 21.8%, 40.9% and 45.7% of the total reads respectively, followed by fungi (9.7%, 4.70% and 3.1%, respectively) and zooplankton (0.3%, 0.3% and 0.1%, respectively), however, the result in JS site was different, in which fungi was dominant (Table 2; Figure 3). Nevertheless, fungi changed to be the dominant in November in all four sampling sits with 10.2% of total reads in TL, 3.8% in FY, 22.4% in WY and 43.7% in JS site, and phytoplankton became to be the secondary assemblages simultaneously (Table 2; Figure 3). Combined with water temperature and salinity condition, the reads proportions of phytoplankton and zooplankton in the three freshwater sampling sites (TL, FY and WY) decreased as temperature drop, while fungi increased in majority of sites except FY. The reads abundance of phytoplankton and zooplankton at the estuarine site, JS, were fewer than the contemporaneous reads of three freshwater sites, and the reads of fungi were also opposite.

Significant spatial differences in values of Chao1 index were deduced (p < 0.05). However, Shannon and Simpson indexes showed no significant spatio-temporal differences. Generally, the three indexes in September at the freshwater sampling sites were all higher than that in November, which was a little different from that in JS sites (Table 2).

Eight samples were divided into two clusters at the 20% level, cluster for JS-A and cluster for the other seven samples, indicating that the microplankton class composition of JS-A had the least similar with that of the other samples (Figure 4(a)), which was also verified by using NMDS ordination method (Figure 4(b)).

We compared the communities of phytoplankton and zooplankton taxa with results obtained from morphological analysis in order to determine potential biases of the primer set used in our study. Species of three phylum and five classes of eukaryotic phytoplankton were annotated by metabarcoding method, which was different from that identified by microscopic examination. Also, more than 104 phytoplankton species were identified by metabarcoding (193 species vs. 89 species) (Table3). Chlorophyceae was the most common assemblage, which was consistent with the finding via morphology, accounting for 12.93% - 79.45% of the total eukaryotic phytoplankton reads, among which Chlamydomonas reinhardtii was dominant, contributing 1.22% of the total reads. Meanwhile, genera with higher reads proportion, such as Chlamydomonas, Chlorella, Crucigenia, Cryptomonas, Actinastrum, Gonium and Dictyosphaerium, were widespread in freshwater of Zhejiang province [27] [28] and were also dominant genera that identified via morphology (TableS1).

For the zooplankton, the comparisons were made with data reported by [29] and [30]. As a widely distributed taxa in Qiantang River, Cyclopoida was identified in this study, and the dominant species of Cyclopoida annotated on were Mesocyclops leuckarti and Acanthocyclops bicuspidatus, which was consistent with the discovery of [29]. However, as another dominant assemblage [30], rotifers were not annotated here.

In this study, we selected ITS with fast evolution and high specificity as the amplicon to describe the community structure and its dynamics of eukaryotic microplankton in Qiantang River lower reaches. Here, the community characteristics of fungi in the Qiantang River were first researched, it’s found that Dothideomycetes, Sordariomycetes, Eurotiomycetes and Tremellomycetes were the most common groups, which were different from that in the Yellow Sea concluded by using the same amplicon [22].

After compared the communities of phytoplankton and zooplankton taxa with results obtained from metabarcoding and morphological analysis, we revealed the same most common assemblage, and discovered 104 more species via metabarcoding. Compared with previous studies [28], it’s speculated that the phytoplankton community had changed somewhat since then, but some dominant specieses maintained unchanged, genera Chlamydomonas and Cryptophyllum were still common dominant groups. Meanwhlie, Anthoathecata, a zooplankton taxa that hasn’t been identified by microscopic examination, was annotated on JS samples. Anthoathecata is an order of class Hydrozoa and distributes on the east and south coast of China. Here, JS sites are located in the estuary area of Qiantang River, indicating a possibility that species of Anthoathecata distribution. However, rotifers were not detected by metabarcoding. The predatory relationship between species may be a considerable reason. Generally, the DNA of the prey does not get separated out during the blast, especially if the prey items belong to the same genus as the predator. In addition, the preference of primer and differences in rRNA gene copy numbers may also explain some of these differences.

Overall, these findings demonstrated that metabarcoding could yield comparable results to conventional methods for several abundant eukaryotic taxa, but that each method has different limitations as far as accurately describing the eukaryotic composition in this river.

Here, geographically, the sampling sites from top to bottom are, in order, Tonglu, Fuyang, Wenyan and Jiashao. There is no hydraulic dam between the four sampling sits, and all of them are in the tidal reach of the river, resulting in a relatively frequent water exchange caused by the flow of tide. Compared with other sites, JS is near estuarine region and more affected by seawater. Correlation analysis showed that there were significant spatial differences in values of Chao1 index (p < 0.05), indicating the species richness had obvious dissimilarity between sampling sites. Cluster analysis and NMDS ordination further confirmed the microplankton class composition of JS-A had the most dissimilarity with that of the other samples, the content of salinity might be one of the reasons [31]. Studies have shown that some spatial differences in eukaryotic plankton α-diversity is more the result of selection by local environmental conditions than dispersal [32], the feasibility of α-diversity based on ITS rRNA gene metabarcoding might be a useful indicator for discriminating ecological condition.