Stratification of microbiomes during the holomictic period of Lake Fuxian, an alpine monomictic lake

In warm monomictic lakes, the hypolimnion is important for accumulating and decomposing organic matter derived from surface production, and the regenerated nutrients will be supplied to the epilimnion through winter vertical mixing. So far, we know little about microbial community composition and function in the hypolimnion when the significant thermal stratification disappears. In this study, we investigated microbial community compositions and functional gene contents by means of metagenomics along a depth profile in the warm monomictic alpine Lake Fuxian during holomictic period. Overall, bacteria were the dominant microbial group at different water depths, while phages had their high relative abundance in the epilimnion. We observed slight thermal but strong chemical stratification even during this typical winter overturn. The anaerobic respiration with nitrate and sulfate as the terminal electron acceptors was accumulated at bottom of hypolimnionin as indicated through metabolic pathway reconstruction. We were able to get 440 metagenome‐assembled genomes (MAGs) and unraveled a high genomic diversity of freshwater pelagic microbiomes along this depth profile. We furthermore defined a new class of “Plancto_FXH1” of Planctomycetes from these MAGs, of which a distinct nitrate reduction operon was identified. Representatives of this phylum mainly thrive in the hypolimnion as previously suspected, but few lineages were detected in the epilimnion. In summary, metagenomics enabled us to find a new group of Planctomycetes, probably involved in denitrification in the hypolimnion in Lake Fuxian, which expand our knowledge on denitrifying bacterial diversity and their denitrification potential in deep freshwater lakes.

Warm monomictic lakes never freeze and are thermally stratified throughout much of the year. The density difference between the warm surface waters (the epilimnion) and the colder bottom waters (the hypolimnion) prevents these lakes from mixing in summer, whereas in winter, the surface waters cool to a temperature equal to that of the bottom waters. Such lakes are widely distributed from temperate to tropical regions (Lewis 1983).
Deep-water winter oxygenation in warm monomictic lakes is vital to the function of lake ecosystems, the lack of which may threaten benthic fauna (Rabalais et al. 2001) and alter biogeochemical cycles (Wetzel 2001).The microorganisms inhabiting the hypolimnion are responsible for accumulating and decomposing organic matter derived from surface production. The regenerated nutrients will be supplied to the epilimnion through winter vertical mixing. It should be noted that the current climate change model is not optimistic for the scenario of thorough mixing in deep lakes, with studies indicating that the ongoing trend of climate warming (Livingstone 2003) may lead to year-round oxygen depletion in deep water within a half century (Danis et al. 2004;Matzinger et al. 2007). Moreover, prolonged stable stratification may be accompanied by increases in hypolimnetic phosphorous concentrations and increased phytoplankton abundances in subsequent years (Lindström et al. 2004;Lavoie and Auclair 2012). The metabolism of organic matter derived from the epilimnion increases oxygen demand, which further deteriorates hypolimnetic oxygen conditions (Naeher et al. 2013). Studies have demonstrated that the ecosystem and biogeochemical cycling in oxygenated hypolimnia are driven by a specific bacterial community that is considerably different from that present in epilimnia (Okazaki and Nakano 2016;Okazaki et al. 2017;Linz et al. 2017). We speculated that oxygen-independent biogeochemical cycling would be predominant in the hypolimnion when oxygen supply was insufficient during the winter holomictic period.
The diversified niches in deep lakes may provide an ideal habitat for the study of speciation and evolution of microbial linkages, especially those "less abundant phyla." In recent years, a combination of improved cultivation techniques and the use of cultivation-independent approaches has led to an increasingly detailed understanding of several groups of abundant and ubiquitous freshwater microbes (Sorokin et al. 2012;Denef et al. 2015;He et al. 2017;Henson et al. 2018;Mehrshad et al. 2018;Neuenschwander et al. 2018;Okazaki et al. 2018). Moreover, ecogenomic approaches have been used to unravel the evolutionary history for some hallmark freshwater lineages (Andrei et al. 2019). Processes involved in altering genomic architecture promoted the transition of sediment/soil-dwelling Planctomycetes into aquatic environments and gave rise to new habitat-specific lineages (e.g., lacustrine-specific). This transition has been well corroborated by the phylogenetic reconstructions of other abundant freshwater bacterial lineages, that is, Betaproteobacteria and Verrucomicrobia (Salcher et al. 2015;Cabello-Yeves et al. 2017). Thus, we expected that deep lakes with a longer water retention time would harbor more habitat-specific lineages of microbes that are specialized to consume relatively recalcitrant dissolved organic matter, especially in the deep hypolimnion.
In this study, Lake Fuxian was taken as a model system to determine whether the annual physical mixing of water can blur the boundaries between epilimnetic and hypolimnetic microbiomes, which promote the development of uniform structure and function in the warm alpine region. Lake Fuxian (24 35 0 N, 102 50 0 E, 1788.5 m a.s.l.) is warm, oligotrophic monomictic lake located on the Yunnan-Guizhou Plateau and is the third deepest lake in China (maximum depth of 158.9 m, detailed information in Supporting Information Table S1) (Wang and Dou 1998). The water volume of Lake Fuxian is ca. 20.6 billion cubic meters, ranking first in freshwater lakes in China. This lake was formed by tectonic faulting, supposedly during Pliocene age. Field sampling was performed during the "theoretical" complete mixing period. Metagenomic deep sequencing was applied to systematically describe the depth profile of microbiomes. Relative gene abundance was used as a proxy for the in situ significance of several core metabolic pathways, which facilitated vertical comparisons among multiple depths.
DNA extraction, library preparation, sequencing, and raw data processing Microbial DNA was extracted from the filters using an E.Z.N.A. ® water DNA kit (Omega Biotek, Norcross, U.S.A.) according to the manufacturer's protocols. Bacterial 16S rRNA gene amplification, sequencing and data processing are available in Supporting Information in the online version of the paper. Metagenomic shotgun sequencing libraries were constructed and sequenced at Shanghai Biozeron Biological Technology. For each sample, 1 μg of genomic DNA was sheared with a Covaris S220 Focused-ultrasonicator (Woburn, U.S.A.), and an average fragment length of approximately 450 bp was used to prepare the libraries. All of the samples were sequenced on an Illumina HiSeq X system using the paired-end 150 bp (PE150) mode. Raw sequence reads underwent quality trimming using Trimmomatic to remove adaptor contaminants and low-quality reads (Bolger et al. 2014). Reads that passed quality control were subsequently mapped against the human genome (version: hg19) and known nonenvironmental bacteria genomes (i.e., Bradyrhizobium, Escherichia, Klebsiella, and Propionibacterium; only complete genomes in the National Center for Biotechnology Information [NCBI] RefSeq release number: 90 was used in this study) using the BWA-MEM algorithm (parameters: -M -k 32 -t 16, http://biobwa.sourceforge.net/bwa. shtml). Finally, uncontaminated high-quality reads were called clean reads and used for subsequent analysis.

Phylogenetic annotation
Taxonomic assignments for the clean reads were made using a customized database that included the genome sequences of bacteria, archaea, fungi, viruses, protozoa, and algae in the NCBI RefSeq database (release number: 90) via Kraken2 (Wood and Salzberg 2014). All of the reads were classified at seven phylogenetic levels (domain, phylum, class, order, family, genus, and species) or unclassified. The abundances of the taxonomic groups were estimated using Bracken, which can produce accurate species-and genus-level abundances, even for multiple near-identical species (Lu et al. 2017).
Metagenomic de novo assembly, gene prediction, and annotation Clean sequence reads were used to generate a set of contigs for each sample using MegaHit with the parameters "-min-contig-len 500" (Li et al. 2016). Open reading frames (ORFs) in the assembled contigs were predicted using Prodigal v2.6.3 (Hyatt et al. 2012). All ORFs were used to generate a set of unique genes after clustering by CD-HIT (parameters: -n 9 -c 0.95 -G 0 -M 0 -d 0 -a S 0.9 -r 1, Fu et al. 2012). The longest sequence in each cluster was considered to be the representative sequence for each gene in each unique gene set. The abundance of each gene within a sample was calculated by using SOAPaligner (https://github.com/ShujiaHuang/ SOAPaligner) and the following formulas: Ab where Ab(S) represents gene abundance; Ab(U) represents single-mapping read abundance; Ab(M) represents multimapping read abundance; and l represents the length of a gene sequence. For each multiple read, there is a species-specific coefficient Co; let us suppose one read in {M} has alignments with N different species, then Co was calculated as follows: We queried the unique gene set of each depth for proteincoding genes based on their assigned identifiers. GhostKOALA tool with GHOSTX searching against a nonredundant set of Kyoto Encyclopaedia of Genes and Genomes (KEGG) GENES was used to assign KEGG Orthology (KO) numbers to each gene in the unique gene set (default parameters; genus_prokaryotes + family_eukaryotes + viruses, Kanehisa et al. 2016). To quantify and compare the abundance of genes involved in carbon, nitrogen, and sulfur cycles, the relative abundance of the detected functional genes was summarized as a proxy of the potential relevance of each pathway in situ.

Metagenome-assembled genomes reconstruction
Metagenome-assembled genomes (MAGs) were constructed for each sample using the following pipeline. (1) Three binning software applications, namely, metaBAT2 (Kang et al. 2015), MaxBin2 (Wu et al. 2014), and CONCOCT (Alneberg et al. 2014) were separately used to perform binning. The final refined bins were obtained using binning-refiner to improve the quality (Song and Thomas 2017). (2) The completeness and contamination of all of the bins was checked using CheckM v.1.0.3 (Parks et al. 2015). Bins with a completeness ≥ 50% and contamination ≤ 10% were considered to be "high quality" and were used in the subsequent analyses. (3) The raw abundance (A j ) of a specific bin in a given sample j was calculated using metaWRAP (Uritskiy et al. 2018) with the following formula: The production length of contig (l i ) and its coverage (c ij ) in sample j was summed across all of the contigs (N) assigned to the bin and then divided by total length of the bin genome. (4) The average nucleotide identity (ANI) between each pair of bins was calculated using the OAT java script (http://www.ezbiocloud.net/sw/ oat) for all of the bins. The bin pairs with > 95% ANI were considered to be the same species in the redundancy step. (5) The phylogenetic annotation of bins was generated using CheckM, where the most similar phylogenetic level of bins was indicated by the comparison between genes of the bins and each set of single copy genes at different phylogenetic levels. (6) All of the genes in a bin were transformed to protein sequences to generate the proteomes for each bin to reconstruct the phylogenetic tree using Phy-loPhlAN (Segata et al. 2013). (7) 16S rRNA genes within each highquality bin were predicted using RNAmmer-1.2 (Lagesen et al. 2007). (8) The phylogenetic tree of 16S rRNA gene sequences was constructed using the maximum likelihood method based on the Kimura 2-paramenter distance mode (Kimura 1980) in MEGA 7.0 (Kumar et al. 2016). In addition, to detect the viral elements at the contig level, VirSorter was used to extract the viral signal from all of the assemblies (Roux et al. 2015).

Functional potential annotation of MAGs
To get an insight on the functional features encoded in the 161 MAGs with completeness ≥ 80% and contamination ≤ 10%, amino acidic sequences predicted by Prodigal v2.6.3 were used as input to EggNOG-mapper to infer functional features based on orthology prediction (Huerta-Cepas et al. 2017). The output file included the annotations for each protein based on Clusters of Orthologous Groups (COG) categories. The abundance of each COG category within a MAG was normalized by the summary of total proteins.

Statistical analysis
Correlation of functional genes and microbial taxa as well with the major environmental factors was analyzed using the R package "Hmisc" (method = "Pearson"). The correlation matrix was reordered according to the Pearson correlation coefficient using the "hclust" method. The R packages "corrplot" and "pheatmap" were used to generate heatmaps. Nonmetric multidimensional scaling (NMDS) analysis was applied to the normalized COG categories of the 161 MAGs, with pairwise cluster distances calculated using Bray-Curtis dissimilarities. The NMDS analysis was performed using the function "metaMDS" in the R package vegan (v.2.2-0), with the number of dimensions set to four after manual inspection of scree plots. Welch's t-test was used to test the hypothesis that two groups of Planctomycetes MAGs have equal means of guaninecytosine (GC) content, genome size, and COG categories.

Sequence data submission
All metagenomic sequence data produced during the study are deposited in the NCBI Sequence Read Archive (SRA) database. All MAGs used in this study can be accessed under the Bioproject PRJNA531348 via the accession numbers SXGV00000000-SYPR00000000.

Results
Environmental profile along the depth profile in Lake Fuxian Although our sampling time in December is regarded as typical winter overturn period of Lake Fuxian, we still observed slight stratification of water temperature, which from the surface to the 50 m was predominantly homogenous, but from 50 m to the bottom there was still a ca. 2 C difference (Fig. 2a). Typically we found quick decrease of DO along the depth profile below 50 m, where the DO was 1.0 mg L −1 at a depth of 100 m and below 0.2 mg L −1 at the bottom. In this context, we defined 50 m as the boundary between the epilimnion and the hypolimnion during the holomictic period.
We also found high concentrations of PO 4 -P, TP, and NO 3 -N but low concentrations of DOC and NH 4 -N in the hyperlimnion (Fig. 2b,c). Based on the correlation analysis between each pair of environmental factors, two groups of environmental factors were formed, one of which included factors such as PO 4 -P, NO 3 -N, TP, and Chl a, which was defined as 'Depth + ENV'; the other group included the factors DO, DOC, Temp, pH, and NH 4 -N, which was defined as 'Depth − ENV' (Supporting Information Fig. S1).

Microbial community composition along the water depth Overview
The metagenomic data set comprised ca. 286 million reads per sample with average length 377 bp. 31.24% of the metagenomic reads were taxonomically assigned based on the analysis pipeline (Table 1). More than 90% of all taxonomically assigned metagenomic reads matched bacteria (90.4-95.6%), with a few representatives observed from archaea (1.19-6.50%), eukaryotes (2.5-4.1%), and phages (0.05-0.47%) (Fig. 3a). Bacteria numerically dominated the genetic composition of the microbial communities, both at the aerobic epilimnion and the oxygen-deficient deep hypolimnia. Archaea were more abundant at a depth of 60 m, below the epilimnionhypolimnion interface (6.50% of total reads), than at the epilimnion (1.19% of total reads) or hypolimnia (3.01% AE 0.47%). With respect to the GC content of the clean metagenomic reads, increasing GC content of microbiomes was observed along the depth profile (Supporting Information Fig. S2).

Archaea
The major archaeal reads were affiliated with phylum Thaumarchaeota (average 67.3%), followed by Euryarchaeota  . Grey point at the right side of (a) indicated the water depth for metagenomic sequencing. Grey shade was used to discriminate hypolimnion from epilimnion.
(average 26.0%), Candidatus Micrarchaeota (average 6.2%), and few additional representatives of Crenarchaeota and Candidatus Korarchaeota. Candidatus Micrarchaeota was only detected in abundance at a depth of 20 m, whereas Thaumarchaeota was more abundant below epilimnion (Supporting Information Fig. S3). At the class level, Halobacteria were extremely predominant below the epilimnion (relative abundance of ca. 80%, Fig. 3c). In contrast, the relative abundances of Thermoprotei (29.33%) and Methanonatronarchaeia (29.87%) were ca. 15-and 600-fold higher in the epilimnion than in the deep hypolimnion, respectively. The abundance of Halobacteria exhibited a negative correlation with the 'Depth − ENV' factor, whereas the relative abundance of Thermoprotei with 'Depth − ENV' showed the opposite trend.

Phages
The maximum number of complete phages occurred at a depth of 20 m (n = 923), decreasing to 203 at a depth of 140 m. Most phages had no classification based on the current annotation pipeline. The uncultured Mediterranean phages (e.g., uvMED, uvDeep, and MEDS5) dominated the assigned phages, followed by Synechococcus phage, Methylophilales phage, and Cyanophage (Supporting Information Fig. S5). Wolbachia phage was only detected in the epilimnion with low abundance.

Functional profile of the microbial communities
An average of 20.38% of the metagenomic reads could be assigned a KO number with putative functions (cutoff e-value 10 −5 ). In this study, we primarily focused on the pathways in carbon fixation, nitrogen, and sulfur cycles along the depth profile.

Photosystem
Overall, the abundances of photosystems I and II peaked at 20 m within the photic zone at the time of sampling (Fig. 4a). In addition, all of the protein subunits of photosystems I and II were present in the genus Synechococcus via metagenomic fragment recruitment, which coincided with their role of oxygenic photosynthesis. Synechococcus is supposed to be the major component of photosynthesis in the winter in Lake Fuxian. A few subunits were sparsely dispersed outside of Synechococcus. For example, both the D1 protein (psbA) and CP47 reaction center protein (psbB) in photosystem II were identified from Deltaproteobacteria.

Nitrogen cycle
Forty-six gene subfamilies targeting eight pathways in nitrogen cycle were recruited from the metagenomic reads. For ammonia fixation, both direct and indirect pathways were abundant in the lake (Fig. 4b). The ATP-dependent glutamine synthetase-encoding gene glnA was the most abundant nitrogen cycle gene detected, especially above a depth of 60 m. In addition, the genes gltB and gltD, encoding glutamateoxoglutarate amidotransferase, also accumulated and can cooperate with glnA in indirect ammonia fixation. Glutamate dehydrogenase (GDH2 and gdhA) exhibited a homogeneous distribution along the depth profile, and this enzyme can directly fix ammonia into α-ketoglutarate in an Nicotinamide adenine dinucleotide phosphate, reduced (NADPH)-dependent reaction. Glutamate synthase gltBD exhibited a synchronous decrease with respect to the decreasing NH 4 -N concentration, whereas the presence of gdhA exhibited the opposite trend.
Processes such as nitrate transport, denitrification, and nitrification intensified in the deep hypolimnion. When mapping all 46 gene families in the nitrogen cycle, the abundances of narGH, norZ, nrtABC, and amoABC significantly increased along the depth profile (Supporting Information Fig. S6). Among the assayed environmental factors, DO exhibited a significant interaction with a specific set of genes. For example, the abundances of nifH and amoBC increased with the decrease in DO along the depth profile (Supporting Information Table S2). In addition, 11 out of 46 nitrogen cycle-related genes showed a positive correlation with the PO 4 -P concentration along the depth profile.

Sulfur cycle
All of the functional genes involved in three steps of dissimilatory sulfate reduction (DSR), namely, the conversion (activation) of sulfate to adenosine 5 0 -phosphosulfate (APS, sat), the reduction of APS to sulfite (aprAB), and the reduction of sulfite to sulfide (dsrAB) were significantly correlated with the PO 4 -P increasing profile along the depth (Supporting Information Fig. S6, Fig. 4c). The assimilatory sulfate reduction, which shares the first step with dissimilatory reduction, was synchronously strengthened, since an accumulation of sat was observed in the deep layers. All of the key enzymes in the Sox system were detected, with the abundance of the gene soxA, a subunit of SoxXA, observed at levels sixfold higher in the deep hypolimnion than in the epilimnion (Supporting Information Table S3). On the contrary, the ATP sulfurylase genes cysD and cysN in the primary sulfate transport complex of Mtb, as well as the cysC gene that encodes APS kinase (in Mtb, cysC is fused to cysN, yielding a bifunctional cysNC gene) was induced by the accumulated DOC in the epilimnion. Similar, not only the enzyme designated DmdA which catalyzes the Dimethylsulfoniopropionate (DSMP) demethylation reaction producing methylmercaptopropionate (MMPA), but also a series of three coenzymes DmdB, DmdC, and DmdD which further metabolize MMPA were abundant in the epilimnion.
The detected bacterial phyla were differentially distributed among the layers (Fig. 5b). For example, one Cyanobacterial MAG that was further assigned to Synechococcus was only identified in the epilimnion, whereas the phylum Nitrospirae was more abundant in the epilimnion-hypolimnion transition area. Acidobacteria only accumulated in the median depth of the hypolimnion (80-120 m), whereas Planctomycetes and Euryarchaeota were primarily distributed in the deep hypolimnion.

Planctomycetes diversity and Candidatus new class
The evolutionary relationships and the taxonomic ranks of the 84 Planctomycetes MAGs were investigated through phylogenomic analysis (Supporting Information Table S5). A maximum likelihood tree (360 Planctomycetes-specific genes) was constructed based on all of the single genomes and metagenomics-recovered representatives published by Andrei et al. (2019) (Fig. 6). Thirty-two MAGs clustered with the newly nominated family Nemodlikiaceae (CL500-3), expanding this clade from 22 to 54 MAGs, with five MAGs occurred within the new families Vodnikaceae and Nixeaceae, respectively (Supporting Information Fig. S10). Excluding the known taxa, a few MAGs formed deep branches within other clades, suggesting distinct potential genomic signatures associated with a specific habitats. Moreover, a monophyletic group with 16 MAGs was distinct from the three known clades anammox Planctomycetes, classes Planctomycetacia and Phycisphaerae. A new class "Plan-cto_FXH1" was nominated based on the bootstrap value and ANI similarity. For the novel class "Plancto_FXH1," the 16S rRNA genes from four out of the 16 MAGs were used to anchor the new clade into the 16S rRNA gene-based phylogenetic analysis, where the MAGs fell within a new clade (Supporting Information Fig. S11a). In addition, an entire canonical nar operon (narGHIJ) was discovered from four MAGs in the new clade, which has far less similarity (ca. 30% similarity) with the two nearest reference genomes (Supporting Information Fig. S11b).

MAGs COG profile
The broad phylogenetic representation of MAGs allowed us to compare the functional potential between taxonomic groups in this ecosystem. NMDS based on normalized counts of functional genes grouped the MAG clusters according to their phylogeny, excepted for Proteobacteria and Planctomycetes (Fig. 5c). Forty-six Planctomycetes MAGs were divided into two groups, Planc_COG_group1 (27 MAGs), containing class Phycisphaerae and new nominated class "Plancto_FXH1," and Planc_COG_ group2 (19 MAGs), containing all the MAGs belong to class Planctomycetacia and four additional unclassified Planctomycetes. The estimated genome size of Planc_COG_group1 (median 2.76 Mb) was significantly smaller than that of Planc_COG_ group2 (median 6.41 Mb) (Welch's t-test p < 0.001). When checking the COG categories, only seven categories demonstrated no significant divergence between the groups, including "A RNA processing and modification," "D Cell cycle control, cell division, chromosome partitioning," "E Amino acid transport and metabolism," "G Carbohydrate transport and metabolism," "H Coenzyme transport and metabolism," "L Replication, recombination and repair," and "V Defense mechanisms" (Welch's t-test p > 0.05). Interestingly, all of the Xing et al. Microbiomes in Lake Fuxian S143 core genome functions were conserved despite large variations in genome size.

Discussion
Deep lakes provide a variety of niches for microbiomes Using a deep sequencing strategy, the results of this study provided an extensive examination of the microbial profile during the annual mixing period in Lake Fuxian. A large number of high-quality MAGs was generated to evaluate the composition and function of the microbial community. The current data set encompassed by far the largest compilation of genomic information available for several freshwater lineages, especially for the phyla Planctomycetes and Verrucomicrobia, which belong to the PVC Superphylum (Wagner and Horn 2006). Both the number of high-quality MAGs and the assembly efficiency were ubiquitous among freshwater metagenomic studies. For example, 38 Verrucomicrobia MAGs were produced from five metagenomes compared to the data set assembly involving 19 Verrucomicrobia draft genomes from 184 metagenomes collected from a eutrophic lake and a humic bog across multiple years (He et al. 2017).
Planctomycetes are ubiquitous bacteria that dwell in many types of habitats but are understudied in freshwater lakes despite being able to provide important ecosystem functions. A few typical freshwater planctomycetal clades (CL500-3, CL500-15, and CL500-37) were nominated years ago (Urbach et al. 2001), but a description of these clades typically stopped at their numerical abundances and phylogenetic diversity. Another bottleneck is a lack of representative axenic cultures for ecophysiological studies. Of 32 validly described planctomycetal species, only three species were isolated from lake water. The study of freshwater Planctomycetes has greatly benefited from the advent of nextgeneration sequencing technologies. Sixty freshwater MAGs branched within two existing classes (Planctomycetacia and Phycisphaerae). Some of the smallest freshwater Planctomycetes genomes have formed new freshwater-specific clades. In this work, we further contributed 84 "high-quality" MAGs to the freshwater Planctomycetes collection. In Lake Fuxian, the estimated genome size of phylum Planctomycetes distributed in a wide range, which indicated cells with large (max. 10.11 Mb) and small genome size (min. 1.98 Mb) coexisted in a habitat with a variety of niches. Although the genome size differed by more than fivefold, core genome metabolism, such as "E Amino acid transport and metabolism," "G Carbohydrate transport and metabolism," and "H Coenzyme transport and metabolism" remained constant. If genome streamlining represents an adaptive evolution from sediment/soil transition to aquatic environments (Andrei et al. 2019), the relative large genomes may indicate massive gene equipment in environments where resources are scarce but diverse and where there is little penalty for slow growth (Konstantinidis and Tiedje 2005), such as the hypolimnion of deep lakes.
The life strategy of Planctomycetes is hot research topic in cell biology and ecology (Fuerst and Sagulenko 2011). The production of secondary metabolites that act as antimicrobials may be one solution to compete with faster growing heterotrophic bacteria. In contrast, a unique cell plan provides a distinct way for free-living Planctomycetes to obtain scarce resource (Wiegand et al. 2018). The results of the latest comprehensive studies indicate the existence of an exceptionally enlarged periplasmic space in planctomycetal cells (Boedeker et al. 2017), where entire high-molecular weight polysaccharides can be transported into their enlarged periplasm. This strategy would allow for the slow degradation of diverse and complex substrates in the sheltered environment of the periplasmic space without the need to secret a plethora of different enzymes. The giant genomes we identified in Lake Fuxian provide evidence that slow growth was also successful in actual competitions.
In this work, based on the assembled MAGs, we were able to identify a monophyletic group formed by 16 MAGs that is distinct to the three known clades anammox, Planctomycetacia and Phycisphaerae. The evidence of a complete canonical nar operon (narGHIJK) with a major facilitator superfamily (MFS) transporter in the new clade may provide a mechanism of surviving in the oxygen-depleted hypolimnion. Anaerobic microbial processes, reactions using oxidized compounds other than oxygen (e.g., nitrate or sulfate) as an electron acceptor, allow microbes to persist in the absence of oxygen. Furthermore, members of the phylum Planctomycetes have seldom been linked to denitrification and/or dissimilarity nitrate reduction, except in one case where Kuenenia stuttgartiensis cells were shown to reduce ammonia via nitrate to nitrite as the intermediate (Kartal et al. 2007). The newly identified clade sheds light on the function of Planctomycetes in the hypolimnion, not only with respect to recalcitrant carbon digestion but also in the nitrogen cycle. This new freshwater Planctomycetes group in denitrification could have been overlooked in previous traditional screening of denitrifying bacterial diversity. Further isolation of strains belong to this new Planctomycetes branch will be helpful for understanding its ecology and physiology at large.

Stratification of microbiomes has occurred since the unvarnished epilimnion-hypolimnion during the holomictic period
This study captured the phenomenon of stratification during the holomictic period of microorganisms in Lake Fuxian with node. An ANI comparison heat map for MAGs of each cluster is shown to the right of the clusters. Three ANI plots were made for the "Plancto_FXH1" plus anammox clade, family Nemodlikiaceae, and the expanded family Nixeaceae. Legends for depths assayed, completeness and a color key for ANI values are shown at the bottom right. respect to both structure and function. The results obtained by metagenomics and 16S rRNA gene amplicon analyses can be mutually supported in the structure dynamics along the assayed depth profile. Measurements of environmental factors indicated that the electron acceptors such as NO 3 -N increased significantly at the bottom of the lake. Meanwhile, Lake Fuxian has an unusual geological background, as the run-off from the surrounding mountains carries large quantities of sulfur into the waters (sulfur concentration in some inlets reaches 37.6 mg L −1 ; Cai et al. 2002). In terms of electron acceptors, chemoorganoheterotrophic can oxidize many different types of organic compounds by transferring electrons from organic compounds to nitrate or sulfate when oxygen is limited, which helps in the degradation of organic materials in the hypolimnion (Knossow et al. 2015;Wenk et al. 2016). The analysis of functional genes and metabolic pathways also supported this hypothesis. A significant increase in the abundance of DSR processes was observed in the bottom water layer. Other processes such as nitrate reduction and denitrification were also activated in the hypolimnion (Llorens-Mares et al. 2015). Combined with the decreased DO in situ, we speculated on the important role of anaerobic respiration in the degradation of complex organic matters.
In contrast to the trend in the decrease of oxygen, TP and PO 4 -P were accumulated in the hypolimnion that further contributes to oxygen depletion. The hypolimnetic microorganisms grow well under conditions with sufficient dissoluble phosphorus. For example, many of the nitrogen cycle-related genes were observed to have a significantly positive reaction with the increasing PO 4 -P in the hypolimnion. In those anoxic environments, fermenting bacteria extracted energy from large organic molecules. The resulting smaller compounds, such as fatty acids and alcohols, were further oxidized by acetogens, methanogens, and the competing sulfatereducing microorganisms (Nakagawa et al. 2012;Bush et al. 2017;Linz et al. 2018). Thus the fermentative short-chain fatty acids can be used as the carbon sources to enhance the biological phosphorus removal under the anoxic condition (Tong and Chen 2007). Moreover, with the hypolimnetic phosphorus escaping to emiplimnion during the vertical mixing, a positive feedback system can be generated in phytoplankton and provide more autochthonous organic matter down to the hyperlimnion that further contributes to oxygen depletion (Kim et al. 2006;Hamilton et al. 2010).
Hypolimnetic oxygen concentration has long been considered to be an important indicator of environmental changes (Müller et al. 2012), since the demand of oxygen in the hypolimnion probably increases during eutrophication. If sufficient amounts of oxygen cannot be supplied to the lake bottom during mixing, the oxygen minimum zone may be extended during the subsequent water stable stratification. A number of hypolimnetic macroorganisms will be either lost or forced into the epilimnion. The hypoxia developed have led to local reduction of mesozooplankton prey in these regions (Vanderploeg et al. 2009). Changes in the structure of food web as well as the microbial loop may further control the microbial community. Moreover, the initial disappearance of oxygen in the hypolimnion can occur prior to any noticeable changes in the productivity of phytoplankton in the epilimnion since the remineralization of organic epilimnetic input by the sediments (Pjevac et al. 2015). The oxygen content of the hypolimnion and the rate of disappearance constitutes a potential "early warning system" of changes in the lake trophic state.

Conclusion
In this study, we focused on microbial community composition and function in the hypolimnetic layers during the vertical mixing. The results indicated that advances in sequencing depth and bioinformatics strategy was highly useful for describing the depth profile of microbiomes associated with the environmental changes. The anaerobic respiration with nitrate and sulfate as the terminal electron acceptors were accumulated at bottom of hypolimnionin, from where the anaerobic phosphate release was transported to the epilimnion thereby providing a potentially significant internal source of phosphorus. The present data set of 440 MAGs encompasses by far the largest compilation of genomic information available for many freshwater lineages. The new class "Plancto_FXH1" of Planctomycetes with a distinct nitrate reduction operon shed light on the function and strategy of "less abundant phyla" in the oxygen-deficient hypolimnion. Furthermore, targeted enrichment and cultivation of microbes would be helpful in solving open questions behind metagenomic analysis, which is important to the association between biogeochemical cycles and their microbial participants.