Article Highlight | 18-Nov-2026

Lakes worldwide harbor 1,370 candidate plastic-degrading microbes, nearly all new to science

Metagenomes from 95 lakes, scaled to 182,661 lakes by machine learning, yielded 4,856 genomes and a culture recipe that isolated a strain removing 36 percent of PET by weight in 90 days

Innovation Press

A plastic bottle that washes into a lake does not stay bare for long. Its surface is colonized, and among the colonizers are microbes able to break the polymer down. A survey in The Innovation suggests they are everywhere: 1,370 candidate species from 95 lakes, 99.9 percent of them new to science.

Lakes are an overlooked sink: unlike the ocean they exchange water slowly and hold less oxygen, so plastic accumulates, and the polymer outlasts the leaves around it. Whether anything digests it has been hard to establish, because earlier surveys inferred degraders from short sequence fragments, leaving most candidate genes unattached to any host genome.

The authors built hidden Markov models around 184 verified plastic-degrading enzymes, paired them with a deep-learning classifier, and searched 715 metagenomes from 95 lakes in 19 countries. It returned 45,154 hits, or 15,715 nonredundant enzyme homologs across 10 polymer types. Assembling the same data yielded 4,856 genomes of mean completeness 82.2 percent, clustering into 1,370 species-level candidates, 609 of them matching no known species. A random forest trained on 37 predictors (R² = 0.73) extended the map to 182,661 lakes larger than one square kilometre.

Two patterns stand out. Plastic waste imports per capita was the strongest single predictor, ahead of climate, and potential was higher at mid-latitudes. Genes for ester-bonded polymers were consistently more abundant than those for polyolefins such as polyethylene, whose carbon-carbon backbone offers a hydrolase nothing to grip.

In 25 lakes with measured plastic loads, potential rose with particle count and levelled off above roughly 7.44 particles per cubic metre. That saturation is a correlation, not a safe limit: past some load, something other than food supply starts to constrain the microbes.

Usefulness is a separate question. Some candidates carry antibiotic-resistance or virulence genes, and a strain that cannot compete in a cold, nutrient-poor lake is of little help. Scoring 1,744 near-complete genomes on degradation potential, ecological risk, adaptation and competitiveness left 11 excellent and 50 high. Because only 3.7 percent of them have cultured representatives, the team reconstructed biosynthesis routes for 41 essential growth factors and designed eight media around the gaps. Plastic left in Xiazhu Lake for six months, then incubated 90 days, produced a community that removed 37.1 percent of PLA and 33.1 percent of PET by weight; the isolate Serratia ficaria HfyG-1 alone removed 35.4 and 36.4 percent, its enzymes sharing under 30 percent identity with known ones yet folding into similar shapes.

The caveats are stated plainly. Everything rests on genes, so the catalog measures putative potential, not measured enzyme activity. It covers fragments of millimetres to centimetres; microplastics below 300 micrometres and nanoplastics were out of scope. HfyG-1 is slower than optimized laboratory strains, and 90 days in a flask is a long way from a lake. What it offers is a map of where to look and a recipe for growing what is found there; for a contaminant that may persist for centuries, that is where the work starts.

Disclaimer: AAAS and EurekAlert! are not responsible for the accuracy of news releases posted to EurekAlert! by contributing institutions or for the use of any information through the EurekAlert system.