The stakes are higher than they might first appear. Across Europe, the ecological health of rivers and lakes is formally assessed under the Water Framework Directive using biological indicators, and in the United Kingdom one of the key tools is the Trophic Diatom Index, or TDI. Diatoms, the single-celled algae with ornate silica cell walls that dominate the phytobenthos of freshwaters, are scraped from biofilm-covered stones, identified either under the light microscope or, increasingly, through high-throughput DNA sequencing. The composition of the diatom assemblage reflects nutrient enrichment and other pressures, and the resulting Ecological Quality Ratio feeds directly into official classifications of whether a river is in good status or failing. But biofilms are far more than diatoms. They are slimy microbial cities, embedding bacteria, fungi, diverse algae, and protozoa within a protective extracellular matrix, and together these organisms drive primary productivity, decomposition, biogeochemical cycling, and pollutant degradation. Scientists have long argued that this broader microbial diversity could yield new diagnostic indicators of human impacts, and recent advances in machine learning and network analysis have made the search for such microbial metrics technically feasible.

The bottleneck has always been data. To identify reliable microbial features that can be upscaled into bioindicators of ecosystem function, researchers need thousands of samples spanning wide geographic areas and meaningful timescales. Generating such datasets from scratch is prohibitively expensive, requiring extensive field campaigns and laboratory analysis. This is precisely where the archive becomes revolutionary. Every year, the Environment Agency of England collects biofilm samples from rivers across the country as part of routine diatom monitoring, and each sample is preserved in an RNAlater-based nucleic acid preservative, a solution of 3.5 M ammonium sulphate, 17 mM sodium citrate, and 13 mM EDTA, before being frozen at -20 °C. A previous study by Baricevic and colleagues in 2022 had shown that such preserved samples remained stable for twelve months, with no significant impact on DNA quality, yield, or the composition of the diatom assemblages. But nobody knew whether the samples would survive intact over multiple years, and the answer to that question determined whether the archive could be treated as a genuine scientific resource or merely as regulatory paperwork.

To find out, the research team selected 64 frozen biofilm samples that had originally been collected and analysed in 2019 and 2020 as part of routine monitoring, chosen to cover a broad geographic spread across England and to have passed rigorous quality control with more than 50,000 reads in the original sequencing run. In December 2022, after two or three years in the freezer, the samples were thawed and reanalysed, and crucially the reanalysis was performed twice on separate sub-samples from each archived specimen. This duplication was the study’s methodological masterstroke, because it allowed the researchers to disentangle any genuine changes caused by storage from the inevitable stochastic noise introduced by sub-sampling, PCR amplification, and sequencing. In total, 192 samples were sequenced across the three sub-sample groups, designated OG for the original analysis and RA and RB for the post-storage replicates, generating 12.9 million quality-controlled reads and revealing 5,138 unique amplicon sequence variants.

The laboratory workflow followed the standard protocol established for routine monitoring. Biofilm-covered stones had been scrubbed with a clean toothbrush in deionised or tap water, the suspension mixed with preservative and couriered overnight to the laboratory, and the samples stored as concentrated pellets at -20 °C. For the reanalysis, samples were thawed, centrifuged at 3000 × g to reform a concentrated pellet, and DNA was extracted using a Qiagen blood tissue kit with an extended overnight lysis step at 56 °C. The target region was the rbcL chloroplast gene, amplified with primers rbcL-646F and rbcL-998R, which capture diatoms but also amplify other non-diatom phytobenthos. The resulting amplicon libraries were purified with SPRI beads, tagged with Illumina Nextera XT unique dual-indexed adapters, normalised, pooled, and sequenced on an Illumina MiSeq. Bioinformatic processing in QIIME2 employed cutadapt for primer removal, DADA2 for denoising and chimera removal, and a multinomial naïve Bayes classifier trained on the diat.barcode reference database for taxonomic assignment, with no filtering so that both diatom and non-target phytobenthic sequences were retained throughout the analysis.

The statistical verdict was unambiguous. At the amplicon sequence variant level, differences in diatom assemblages were overwhelmingly explained by differences between sample sites, accounting for 38.3 per cent of the variation, while storage conditions between the original and post-storage sub-samples explained less than 1 per cent and were not statistically significant. When the researchers specifically compared two versus three years of storage, the result was equally insignificant. At the genus and species levels, no taxa were detected at significantly higher abundance after storage, and the vast majority of genera, 84.0 per cent, were found in all replicate groups. The few genera and species that appeared unevenly between replicates were detected at very low levels, typically fewer than 1,000 reads across only one to three samples, and almost all belonged to genera that were more widely detected across the dataset. Only two diatom species, Surirella solea and Placoneis constans, were uniquely detected in post-storage replicates without their genus being more widely found, and even these appeared at negligible abundance. Curiously, the non-target species Quercus robur, the English oak tree, was detected in three post-storage replicates only, almost certainly the result of trace contamination or incidental amplification rather than any meaningful biological signal.

Perhaps most importantly for regulators, the ecological status metrics derived from the DNA were remarkably consistent. The mean difference in TDI values before and after storage was just 3.18, and Pearson’s correlation coefficients between the three replicate groups ranged from 0.892 to 0.980, all highly significant. When ecological status classes were assigned, 75.8 per cent of samples received the same classification across all replicates, and 79.0 per cent were either identical or differed by only one class. In pairwise comparisons, the original and first post-storage replicate agreed on the same class for 90.4 per cent of samples. Only four samples stood out as clear outliers, with differences greater than one class depending on which replicates were compared. In one of these, the post-storage replicate contained a much larger proportion of the diatom Amphora pediculus, at 47.1 per cent, than the original and first replicate, at 22.2 and 17.5 per cent respectively. But because no species were differentially abundant between replicate groups across the dataset as a whole, the researchers concluded that these outlier differences were not caused by fundamental changes from storage, but rather by the same sub-sampling and PCR variation that plagues any routine monitoring data.

This interpretation places the study in a broader scientific context. The magnitude of the insignificant differences observed between original and post-storage analyses was similar to that reported in previous work when comparing differences between different analysts and different sequencing instruments, suggesting that the noise inherent in the analytical pipeline itself is comparable to or greater than any effect of multi-year storage. Comparable findings have emerged from studies of other microbial sample types. Researchers examining bacterial communities in faecal samples stored for five years in preservative at -80 °C found them comparable to reference samples, and multiple shorter-term studies have reported similar stability. However, the literature is not entirely uniform, as one study found that RNAlater significantly affected bacterial communities compared to liquid nitrogen freezing at one of two sites tested, while finding no significant differences for fungal communities. The authors of the current study caution that extrapolating their results to the wider microbial community requires care, because research on the stability of bacterial communities by 16S metabarcoding is contradictory and difficult to compare directly, and little research has been conducted on the impact of storage conditions on fungi or protists within biofilms.

The practical implications extend well beyond diatoms. The study further evidences RNAlater-based preservation of freshwater biofilms as a viable alternative to ethanol, one of the standard recommended methods under European standardisation guidelines. For logistical, health, and safety reasons, the Environment Agency has restrictions on the use of ethanol, and therefore sought alternatives for the sampling and storage of biofilm samples. The confirmation that these preserved samples remain stable for up to three years means that regulators can now maximise the value of their historic raw biofilm samples, reusing them for purposes far beyond their original intent, including monitoring other non-microbial taxonomic groups and exploring the responses of broader microbial communities to environmental change. What was once a freezer full of completed regulatory assessments has been recast as a living archive, ready to yield the large, cost-effective datasets that will underpin the next generation of microbial metrics for freshwater biomonitoring, and with them, a far richer understanding of how river ecosystems respond to the mounting pressures of a changing world.

Subject of Research: Environmental DNA

Article Title: Influence of storage time on the stability of diatom assemblages using DNA from riverine biofilm samples

Article References: Warren, J., Butler, S., Evens, N., Hunt, L., Kelly, M., Newbold, L., Read, D. S., Taylor, J. D., & Walsh, K. (2024). Influence of storage time on the stability of diatom assemblages using DNA from riverine biofilm samples. Metabarcoding and Metagenomics, 8, Article e129227. https://doi.org/10.3897/mbmg.8.129227

Image Credits: AI Generated

DOI: 10.3897/mbmg.8.129227

Keywords: environmental DNA, diatoms, biofilm, biomonitoring, metabarcoding, river health, Influence, storage, time, stability, diatom, assemblages