Two lakes can sit near the same city, share similar shapes and depths, and still host radically different microbial worlds. In this study, the strongest dividing line was how nutrient-rich those lakes had become. Over the last decades, lakes have been particularly affected by human activities, species invasion, increased surface temperatures and heat waves associated with global change.

These pressures add to natural seasonal variations and enhance eutrophication, ultimately leading to major changes in lake ecosystem functioning worldwide. Microbial communities are key contributors to ecosystem functioning, quickly reacting to disturbances, and are therefore often investigated to assess lake health.

Eutrophication is a major driver of these communities, promoting both phytoplankton, including cyanobacteria, and heterotrophic bacteria. But its impact is often hard to distinguish from lake morphology, land cover and uses, while community variation over time also needs to be accounted for.

Existing time series often include only a limited number of lakes, and very few functional comparisons between trophic statuses are available. The study asks how trophic status is correlated with microbiome structure and functional potential during summer, when primary production peaks.

The hypothesis is that different trophic statuses harbor different microbial communities in taxonomical and functional composition, and that higher trophic status induces greater temporal variability during summer. The Greater Paris offers a natural laboratory because its lakes occur across a limited geographical area and have few confounding factors in climate, geological context, area, depth and pH.

Nine shallow lakes with comparable morphologies were sampled monthly over the 2021 summertime, using 16S rRNA gene amplicon and shotgun metagenome sequencing. Nine lakes were surveyed monthly from June to September 2021: Jablines, Vaires-sur-Marne, Cergy large, Cergy small, Créteil, Bois-le-Roi, La Grande Paroisse, Champs-sur-Marne and Verneuil-sur-Seine.

They were located within about a seventy-kilometer radius around Paris and selected for similar area, depth and absence of stratification. These lakes were former sand and gravel quarries transformed into human leisure centers between the nineteen sixties and the nineteen eighties.

In each lake, the water column was sampled at three mid-lake locations, labeled W1, W2 and W3, to account for spatial heterogeneity. At each location, five liters were collected at three depths with a Niskin bottle and pooled in equal volumes to form a depth-integrated sample.

A total of one hundred five samples were collected, although Cergy large could not be sampled in June. Filtration was performed on site within one hour, and other subsamples were pre-filtered through a fifty-micrometer mesh to remove large particles.

Chlorophyll a, a proxy of phytoplankton biomass, was measured from five hundred milliliters of water filtered onto zero point seven micrometer filters in triplicate by spectrophotometry. Trophic status was defined from Carlson’s trophic state index ranges: oligotrophic below two point six micrograms per liter, mesotrophic from two point six to seven point three, eutrophic from seven point three to fifty-six, and hypereutrophic above fifty-six.

Phytoplankton composition was determined visually, with taxonomic identification and relative cell counts based on two hundred to four hundred random individuals per sample. For each depth-integrated water column, between one hundred fifty and two thousand milliliters of pre-filtered water was passed through zero point twenty-two micrometer membranes, and the filters were flash-frozen in liquid nitrogen.

16S rRNA gene amplicon sequencing was performed on all one hundred five samples. Shotgun metagenomics was performed on thirty-five samples from W2, while the 16S rRNA gene sequencing covered W1 to W3 across the sampled lakes and dates, except for Cergy large in June.

Total DNA was extracted with the PowerLyzer PowerSoil DNA extraction kit, including a prior bead-beating step. Two extraction-blank controls were performed and incorporated into the 16S rRNA gene amplicon sequencing analyses. Genomic DNA from each of the thirty-five W2 samples was sequenced with Illumina MiSeq, yielding fifteen point eight million paired-end reads per sample on average.

Reads with a phred score below twenty were discarded, and human-associated reads were removed based on the GRCh38 human genome assembly. The samples were assembled individually using SPAdes, and contigs were annotated taxonomically and functionally using CAT and EggNOG-Mapper.

The final dataset consisted of seven thousand nine hundred ninety-four annotated KOs, or KEGG Orthologies. Taxa and gene content were summarized with richness, evenness and Shannon indexes, while KOs and ASVs present in one lake throughout all four sampled months defined its core gene and taxa contents.

Month-to-month turnovers were computed for each lake. A linear mixed model tested whether turnover differed among trophic statuses while accounting for month comparisons and lake-specific effects. Using average chlorophyll a over four months, two lakes were mesotrophic, five were eutrophic and two were hypereutrophic.

Photosynthetic eukaryotes dominated phytoplankton in the two mesotrophic lakes, one eutrophic lake and one hypereutrophic lake. That pattern shows that whether eukaryotes or prokaryotes were the main chlorophyll a producers was not a function of trophic status. The two hypereutrophic lakes also differed: Verneuil-sur-Seine was dominated by Ceratium, while Champs-sur-Marne was dominated by Cyanobacteria.

In panel C, the stacked monthly bars show substantial within-lake shifts across the four sampling months: for example, the pale prokaryote fraction expands markedly in LGP and CSM, while it remains a narrow strip in JAB and VSM. The map in panel A also shows all sites clustered within the Paris region, rather than spread across the country.

The thirty-five metagenomes yielded seven thousand nine hundred ninety-four unique annotated prokaryotic KOs, with fifty-two percent shared between all lakes and all dates. All lakes showed similar gene-content richness and Shannon diversity, while each lake’s core gene-content consisted of about five thousand one hundred thirty-eight KOs.

Those core KOs represented most of the richness and ninety-nine point eight percent of KO abundance for a given lake. Low month-to-month KO turnover confirmed that the core gene-content was dominant and stable throughout summer, whatever the lake and its trophic status.

Taxa showed greater differentiation: only five of five thousand five hundred fifteen unique ASVs were present in every lake and on every date. The most abundant phyla were always Actinobacteriota, Cyanobacteria, Bacteroidota, Proteobacteria, Planctomycetota and Verrucomicrobiota.

The core taxa-content accounted for a higher fraction of ASV richness in mesotrophic lakes than in eutrophic and hypereutrophic lakes. The core taxa-content represented seventy-nine point five percent of relative abundance in mesotrophic lakes, sixty-two point one percent in eutrophic lakes and thirty-four point three percent in hypereutrophic lakes.

Month-to-month ASV turnover was significantly lower in mesotrophic and eutrophic lakes than in hypereutrophic lakes: fifty-three versus sixty-three and eighty-three percent, respectively. The lake polygons slightly overlapped, but VSM and JAB clustered apart from the others, while VSS displayed by far the largest polygon.

Segregation was significantly explained by trophic status, while no difference was observed according to month or spatial distance between lakes. Intra-summer heterogeneity was significantly higher in hypereutrophic lakes than in eutrophic and mesotrophic lakes, which displayed similar variability.

Heterogeneity was particularly high for VSS, as illustrated by its large polygon area. In panel A, the large circles show that core genes account for roughly seventy to eighty percent of gene reads across lakes, whereas the smaller ASV circles are much more variable, from about three to eleven percent.

Panel C shows a shared phylum-level makeup, but the stacked bars visibly shift in the relative contributions of Cyanobacteriota, Actinobacteriota, Bacteroidota, and Proteobacteria across lakes and months. Month-to-month Bray-Curtis dissimilarities for gene and taxa contents were significantly correlated, with a Spearman rho of zero point sixty-five.

The two hypereutrophic lakes, CSM and VSS, displayed the highest gene-content and taxa-content dissimilarities, while eutrophic and mesotrophic lakes did not differ significantly. BLR displayed the lowest ranges of both gene-content and taxa-content dissimilarities.

In the scatterplot, the blue and red points cluster toward the lower-left, while the green points occupy much more of the upper-right region. The orange points sit mostly between these groups, showing how the lake categories are distributed along both dissimilarity axes.

To examine functional potential related to major biogeochemical cycles, twenty-eight BGC marker genes were selected, together accounting for zero point seven percent of total KO relative abundance. The functional potentials of lakes were segregated according to trophic status along the second PCoA axis.

Two genes involved in polyphosphate synthesis and hydrolysis, ppk1 and ppx, together explained sixty-nine point eight percent of the mesotrophic-to-hypereutrophic comparison. Their relative abundances were higher in mesotrophic lakes than in hypereutrophic lakes.

Both genes were mainly detected on contigs affiliated with Actinobacteriota. No difference in nitrogen metabolism marker genes was detected, while psbA and rbcL were respectively sixfold and fivefold more abundant in BLR than in other lakes.

In panel A, the colored lake polygons occupy partly distinct regions, with the red BLR samples extending farthest to the right along axis one. In panel B, the pie chart is dominated by the ppk1 slice at fifty-four point six percent, followed by the non-significant portion at twenty-nine point six percent and ppx at fifteen point two percent.

Comparable lakes near Paris, with trophic status ranging from mesotrophic to hypereutrophic, showed that trophic status affected community structure and functional potential in summer. Functional potential was much more stable than taxa composition within each lake, and most of it was shared among all lakes.

Whether the driving effect and hypothetical regime shift in hypereutrophic lakes continues during winter and spring needs to be tested. One eutrophic lake displayed very stable communities compared with the others, suggesting that trophic status alone cannot explain every observation.

Peri-urban areas such as the Greater Paris provide excellent settings because lakes with varied trophic status occur in close vicinity. Across nine Paris-area lakes, trophic status shaped both prokaryotic community structure and functional potential, while gene content stayed relatively stable.

The most eutrophic lakes were the most heterogeneous over summer, pointing toward a possible microbial regime shift.